East African University Rwanda Scientific Journal (EAURSJ)
Socioeconomic Determinants of Climate-Smart Agroforestry Adoption in Rwanda
Donatien Ntawuruhunga1*, Enock Nyisingize1, Marie Therese Mukanyangezi2, Mathias Twahirwa3
1Department of Business Administration, East African University Rwanda, P.O. Box 179, Nyagatare, Rwanda
2 Faculty of Business and Development Studies, Kibogora Polytechnic, P.O. Box 50, Nyamasheke, Rwanda
3 Faculty of Theology and Religious Studies, Protestant Institute of Arts and Social Sciences, P.O. Box 619, Huye, Rwanda
*Correspondence to: Department of Business Administration, East African University Rwanda, P.O. Box 179, Nyagatare, Rwanda. E-mail address: drdonatien@eaur.ac.rw (Donatien Ntawuruhunga), ORCID: https://orcid.org/0000-0002-2945-129X
Abstract
Climate-smart agroforestry (CSAF) represents a novel approach to modern farming, yet research has not thoroughly explored the complexities of CSAF practices, particularly the factors influencing their adoption. By employing an advanced methodology, the “Multivariate Probit” (MVP) approach, this study aimed to explore bundles of CSAF practices and their complex interplay with farmer characteristics (demographics, assets, institutions) by going beyond single-practice adoption. The study was driven by two research questions: (i) What are the factors driving the adoption of CSAF practices in the Bugesera and Rulindo agroecological zones of Rwanda? (ii) How do CSAFs correlate and complement each other, guiding more targeted interventions? A sample of 381 farmers was selected to acquire data, and a survey was conducted based on agroecology from Bugesera and Rulindo (Rwanda). The seven CSAF practices defined the outcome variables (home garden, alley cropping, improved fallow, multipurpose trees, plant crop combination, shelterbelts/windbreaks, and silvopasture). CSAF practices were found to be complementary to each other, which implies that a farmer’s adoption choice of a given CSAF practice was conditional on the adoption of the others. Geographic area (AEZ), altitude, gender, age, farm size, farmer’s experience, radio, livestock size, and farm – river distance were significant factors influencing the adoption of CSAF practices. More importantly, when farmers decide whether or not to adopt CSAF, their age, the size of their farms, and the length of time they have been farming are the most significant factors influencing their decision to implement these practices. These results have an important implication: highlighting the importance of targeted interventions to enhance CSAF adoption, including physiological, climatic, sociological, and economic variables, as well as CSAF practice options. Taking these factors into consideration can create an enabling environment for CSAF adoption, boosting climate-resilient farming, and sustaining agricultural productivity.
Keywords: Climate-smart agroforestry; Adoption determinants; Rainfed subsistence farming; agriculture business; Multivariate Probit regression
- Introduction
In recent years, the forestry industry has undergone significant evolution. This economic sector has shifted its focus beyond timber production to embrace modern technologies. The current forest husbandry approach emphasizes managing trees for diverse purposes to meet social, economic, food security, and ecological needs while addressing the effects of climate change. Climate-smart agroforestry (CSAF) represents a cutting-edge development in the forest industry and research, using an integrative approach to land husbandry and crop-tree integration for various purposes (Glover et al., 2013). Climate change seriously threatens global agriculture (Shahbaz et al., 2022). The effects of climate change have increased awareness of CSAF’s potential to enhance agricultural productivity, income, and food security in rural areas where most people live and rely on rainfed subsistence farming. Literature indicates that prominent scholars increasingly recognize CSAF’s role in preserving biodiversity and promoting inclusive and sustainable agricultural growth while tackling climate change (Harvey et al., 2006; Shackleton et al., 2012). This rapid paradigm shift toward a more integrative farm management approach underscores the need to promote inclusive and sustainable agricultural growth despite the effects of climate change. CSAF is an ideal farming intervention for transforming rural landscapes into sustainable, productive lands that contribute to rural development gains (Patel et al., 2024).
In Sub-Saharan Africa (SSA), agrarian production has been hampered by land degradation and the shortage of quality irrigation water induced by climate change effects (Raj, 2017). This has resulted in a vicious cycle of poverty and hunger. CSAF is one of the innovative and sustainable farming practices that integrates crops with trees to increase and diversify food productivity. CSAF is an important farming option to secure food security for most people in rural areas who rely on subsistence farming (Venkatesh et al., 2014).
CSAF has emerged as a novel and vibrant research field and development over the last 20 years in the wake of the Millennium Development Goals (MDGs) and its continuation with the current Global Goals (SDGs). CSAF provides diverse services to rural households and biodiversity conservation while adapting to climate change. Focus has been shifted to this field by researchers, scientists, scholars, and experts in agriculture, forestry, and nature conservation (Raj, 2017). Studies in this field apply scientific principles to address challenges facing agricultural production and the environment, thereby jeopardizing food security and economic stability (Jama et al., 2006). CSAF farming generates diverse utilities such as fruits (home consumption), fuelwood (culinary), fodder (cattle-housed for zero-grazing), standing trees (stakes for climbing crops), logs, and poles/timber meant for commercial purposes. In the face of climate change, CSAF helps farmers eliminate production risks and crop failure and meet their basic needs (ICRAF, 1993).
In the face of the adverse effects of climate instability, smallholder farmers have been modernizing their systems by integrating trees with crops and/or animals to sustain production in many parts of the world. These common CSAF practices of smallholder farmers in the face of climate challenges include taungya, home gardens, scattered trees on farms, improved fallows, and windbreaks, among others (Awazi et al., 2021). CSAF is one “climate-smart agriculture” added approach as a joint effort to improve agricultural production to meet the needs of the world population, degrading arable lands, and biodiversity (Ntawuruhunga et al., 2023). In developing countries, CSA is essential for economic development, food security, environmental sustainability, and the preservation of ecosystems. Diverse practices of CSA can vary following adaptation to geographical area and climatic conditions (Shahbaz et al., 2022). Climate-induced shocks and other disturbances threaten agricultural productivity (Amadu et al., 2020).
While CSAF holds significant promise, its adoption has been uneven across landscapes (Glover et al., 2013). Despite successful technological advancements, rural development initiatives regarding trees on farms have failed due to neglecting socio-economic factors (Mekuria et al., 2004; McNeely and Schroth, 2006; Cerdán et al., 2012). Projects failed due to the neglect of producing concrete benefits for farmers. Realizing the need for success in the forest industry, academic and research institutions specializing in CSAF are urging increased research on socioeconomic patterns (Mekuria and Waddington, 2004). Planners for rural investments in rural areas now acknowledge the vital role of socioeconomic factors in CSAF adoption. (Montagnini, 2006) emphasized that successful farming with CSAF requires a thorough assessment of socio-economic factors while analyzing the feasibility of CSAF systems and factors influencing their adoption.
Rural transformation is happening everywhere for sustainable agricultural growth, but not everyone benefits (Sunderland et al., 2017). Regarding the growing population living in chronic poverty, the question remains of who benefits, how, and why (Rahman et al., 2014). If marginal and abandoned land is converted into cultivation through CSAF, it could bring multifold benefits to the rural households who rely on rainfed subsistence farming (Patel et al., 2024).
As a low-input, climate-smart farming tool, CSAF enhances food production and economic stability in sub-Saharan Africa (SSA) (Garrity, 2012). It integrates traditional agroforestry (AF) farming approaches with climate adaptation strategies. It is gaining attention and interest in academia, industry, and research institutions due to its edaphic, ecological, economic, and social patterns (Ndoli et al., 2021). CSAF is increasingly recognized as a cost-effective solution for food shortages, famine, climate shocks, soil regeneration, landscape restoration, and improved food security (Mbow et al., 2014). Food security and economic stability are primary motivators for CSAF adoption in SSA, particularly in rural areas where most people rely on rainfed subsistence farming (Brown et al., 2013).
Trees in CSAF farms are a keystone of farm productivity. Trees can transform landscapes, providing multiple benefits (utilities) for ecosystems and farming communities (Xu et al., 2013). Xu et al. (2013) classified trees on farms, such as CSAF, into four criteria (Table 1).
Table 1. CSAF classification
| Classification | Description |
| Structural | System composition and spatial-temporal arrangement impact diversity, sustainability, and productivity. Introducing woody species across various farm niches enhances these outcomes. |
| Functional | Encompassing the tree component’s role, like timber, fruit, fodder, and medicine, increases the system’s product variety and serves as a safety net for farmers. Tree services, like windbreaks and erosion prevention, are also valuable. |
| Socioeconomic | Defines the system’s purpose for human livelihoods, often categorized as subsistence, commercial, or intermediate production. CSAF may target social goals like poverty alleviation and food security. |
| Ecological | For CSAF system suitability in various environments, tailored for tropical, temperate, and arid regions, considering environmental, ecological, and biological conditions. |
Source: Adapted from Xu et al. (2013)
Today, climate shocks negatively impact livelihoods and hinder economic growth (Ong and Swallow, 2003; Brown et al., 2018). Rwanda’s agriculture is the backbone of its economy. This sector employs 3.4 million farmers and contributes 23% of its gross domestic product (GDP), with 69% of households engaged in farming (NISR, 2023). Given the high population density, natural and planted forests are insufficient to meet the growing demand for forest products. CSAF is a lasting solution to this long-standing challenge (Rutunga et al., 2007; Ndayambaje and Mohren, 2011).
The world population growth is unmatched by production. This suggests that people will increasingly consume more resources under the growing constraints linked to climate change. We must cope with climate change shocks to meet food security needs while maintaining our natural ecological systems. CSAF can help increase productivity and build resilience. Congruently, CSAF is a smart farming approach that involves mixed crop-tree farm production to raise productivity and cope with and repair damages induced by climate change. Scientific evidence shows that CSAF farming is vital in tropical agriculture for forest conservation, food production, and environmental and climate change resilience (Rahman et al., 2007; Leakey et al., 2012). CSAF is a source of multiple benefits in rural areas where most people, such as subsistence farmers, unemployed, and poor landless farmers, survive (Nath et al., 2005).
However, CSAF farming struggles to gain adoption in Rwanda due to limited knowledge, scarce resources (often on marginal land), and insufficient technical skills required for adoption. Despite the Indigenous trees being kept on farms for various uses, farmers in the study area face challenges in implementing CSAF sustainably due to unpredictable weather patterns, drought, and erratic rainfall (Oduniyi and Tekana, 2019). Notably, farmers in the eastern savannah dry lowlands, traditionally reliant on pastoralism and rangeland, lag in adopting climate-smart farming practices. In contrast, the rural highland area of Rulindo embraces a tradition of crop farming.
If Rwandan agriculture has to flourish, building on its proximity (Kigali-Bugesera (44.1km), Kigali-Rulindo (36.7km) to the City of Kigali (a market for farm commodities), the situation in Bugesera and Rulindo has to improve in all sectors to promote inclusive and sustainable agricultural growth. Despite the challenges induced by climate change, CSAF can positively and significantly impact livelihoods with the desired level of diversification and sustainability (Dwivedi et al., 2007). Modern farming technologies of farm-industry linkages have revolutionized systems to be more sustainable than subsistence farming (Saxena, 2000; Kareemulla et al., 2005).
In Rwanda and tropical SSA, diverse initiatives aim to boost smallholder productivity, with varying success. Rwanda’s mountainous geography and varied precipitation create five distinct agroecological zones (AEZs): East-Rwandan lowlands, Temperate central highlands, South-Rwandan humid mountains, Kivu-sea, and North-Rwandan dry mountains. Each zone corresponds to specific agricultural practices, associating particular plants, crops, and pastoralism. CSAF tree species, tailored to withstand dry spells, demonstrate adaptability to zones, such as the east savannah dry lowlands plagued by termite infestation. CSAF adoption hinges on biophysical, climatic, economic, social, and cultural conditions and policy involvement.
Even though various studies have given attention to the association between CSAF and smallholder farm productivity using regression such as binary logistic regression analysis (e.g. (Mukundente et al., 2020; Amare et al., 2018), general linear model (Jagoret et al., 2012; Coulibaly et al., 2013; Mercer and Frey, 2014; Armengot et al., 2016), majority of these studies, assessed the level of productivity as a function of agricultural technology, input use, farm size, and good agronomic practices or animal husbandry, among other variables, using various statistical models, with emphasis on linear regression. A research gap exists on factors determining farmers’ adoption/non-adoption choices of CSAF.
Therefore, this study focused on how different factors influence farmers’ choices to adopt integrated tree-crop systems in unique zones (Bugesera’s dry lowlands versus Rulindo’s temperate highlands) using advanced statistical modeling (Multivariate Probit for correlated decisions) to understand the complex adoption drivers and impacts on livelihoods. The study was driven by two research questions: (i) What are the factors driving the adoption of CSAF practices in the Bugesera and Rulindo agroecological zones of Rwanda? (ii) How do CSAFs correlate and complement each other, guiding more targeted interventions? The study is scientifically relevant, highlighting key variables (CSAF, Rwanda’s agroecological zones, CSAF impacts, and specific statistical modeling, such as MVP), and indicating research on contextual drivers, knowledge, and socio-economic outcomes, making it highly relevant for sustainable agriculture studies.
The study introduces “Climate-Smart Agroforestry (CSAF)” as a timely topic aligned with the Sustainable Development Goals (SDGs) and crucial for improving soil health, enhancing climate resilience, and promoting food security, thereby advancing global sustainability goals. The study highlights contextual differences by comparing Bugesera (a semi-arid region) and Rulindo (a highland region), revealing how agroecological zones, climate, and demographics influence adoption. The study employs the “Multivariate Probit” (MVP) approach, which is ideal for modeling simultaneous decisions to adopt multiple CSAF practices and for capturing unobserved correlations (e.g., adopting fruit trees alongside soil conservation). The study addresses key research gaps: it goes beyond single-practice adoption to explore bundles of practices and their complex interplay with farmer characteristics (demographics, assets, institutions). Finally, the study informs policy decisions: findings inform targeted interventions to promote CSAF, benefiting smallholder farmers in Rwanda and similar contexts.
Rwanda’s agriculture remains dependent on smallholder farmers with limited technology. Ironically, CSAF farming has proven to be a simple, resourceful, affordable, and efficient climate change mitigation strategy. As a new approach in modern farming systems, there is limited literature on various CSAF practices and their determinants of adoption. Using data from 381 farmers in Bugesera and Rulindo agroecological zones, this study necessitates a detailed and nuanced analysis of complex adoption dynamics in a vital agricultural region, making it a compelling area of investigation.
- Materials and methods
2.1. Study area
Study areas fall into two contrasting agroclimatic zones of Rwanda ─ the Bugesera area in the eastern semi-arid savannah lowland zone and the Rulindo area in the temperate zone of the central highlands. Besides their agroecological disparities, these two sites were purposively selected for investigation based on their geospatial characteristics, biophysical and climatic conditions, and the diversity of patterns of CSAF in terms of composition (species diversity), planting geometry and configuration (spatial niches), and production (Figure 1).
Bugesera area is located between Latitudes 2009’S and Longitude 30005’E, with a total geographic area of 1,337 sq km (Figure 1). Owing to its low altitude and forest deforestation, with scant rivers and streams and a deficiency of precipitation, Bugesera is a drought-prone region. The average temperature oscillates between 26 and 29°C. Its altitude varies from 1,100 to 1,780m above the mean sea level. The average annual rainfall is 943mm. As per the Bugesera District Development Strategy (DDS 2018/19-2023/24) (2019), the flora of this area is remarkably dominated by two plant formations: the dense shrubs covering the hills and the grassy savannahs covering the dry valleys and the slopes of the hillsides. Also, the vegetation comprises acacia trees, euphorbia, and cactuses intertwined with gramineous and spiny bushes. Further, the vegetation encompasses other rare species that are not gigantic but intertwine with the bushes and trailing lianas. The area is also dominated by undulating hills and mid-hills alternating with dry valleys and marshlands alongside large rivers (ibid.).
Located in the temperate zone of the central highlands, the Rulindo area is mostly characterized by mountains with rivers and streams flowing through a succession of valleys. It has an area of 567 sq km (Figure 1) with a tropical climate characterized by alternating precipitation seasons and dry spells. The geographical coordinates of this area comprise Latitude 1o44’S and Longitude 29o59’E. As per the Rulindo District Development Strategy (DDS 2018/19-2023/24) (2019), the vegetation is largely composed of food crops on hillsides and valleys with woodland eucalyptus, grevillea, calliandra, and cyprus also co-existing. It also harbors ferns, the eragrostis (ishinge), a sign of soil degradation (ibid.). The mean annual temperature is 19°C., while mean rainfall normally reaches 1,243.3mm p.a.
Figure 1. Map of Rwanda with study areas (Bugesera and Rulindo) (adapted after CGIS)
2.2. Research design and approach
Research strategies and choices, completed with time horizons (Lankamo & Ramalingam, 2025), generally entail and reinforce design decisions in a research project. Concerning the time horizon, this study relied on a cross-sectional design, which denotes that cross-sectional datasets were collected from farmer households, community leaders, farmer groups, and government officials at one point. With a mixed methods (pragmatic) approach, descriptive and explanatory sequential research designs were utilized in this study.
2.3. Source of data and study population
This study integrates both empirical and secondary data sources. Empirical data were collected from various stakeholders, with a focus on smallholder farmers. Secondary data were sourced from official reports, published literature, and policy documents. Demographic and socioeconomic data were accessed from the National Institute of Statistics of Rwanda (NISR). The study population comprises rural smallholder farmers in two selected districts — Bugesera and Rulindo, which were chosen for their representation of contrasting agroecological zones. According to the National Institute of Statistics (NISR, 2021) report, Rulindo district hosts a population of 204,000 comprising 84,000 smallholder farmer households, while Rulindo district hosts a population of 160,000 comprising 82,000 smallholder farmer households. Data collection utilized both quantitative and qualitative procedures.
2.4. Sampling design and sampling procedure
The study used a combination of probability and non-probability sampling methods to achieve its objectives. A probability sampling method, which ensures that every unit in the population has an equal chance of selection for a sample, was used for collecting quantitative household data. Concurrently, non-probability sampling targeted specific groups (Lankamo & Ramalingam, 2025), such as extension officers, community leaders, and farmer groups, based on predefined criteria. In this study, smallholder farmer households were designated as the primary sampling unit, as climate-smart agroforestry is adopted at the household level, and outcomes, such as income and food security, are often most evident at the household level. Also, qualitative methods were employed to capture broader CSAF adoption and local impacts.
A multi-stage sampling procedure was applied, progressing from broader to narrower samples involving stratified and simple random sampling techniques. In the first stage, the two regions described above were purposively selected based on their agroecological zones (Table 2) and the diversity of attributes of CSAF in terms of composition (species diversity), planting geometry, configuration (spatial niches), and production. In the second stage, the two regions were stratified using their elevation: the lowland (Bugesera) zone and the highland (Rulindo) zone. In the third stage, a list of villages was generated in each zone. The listing of villages relied on information from extension officers about agricultural potential, accessibility, and the level of CSAF practices and production. In the fourth stage, lists of households were produced in each village.
Table 2. Study sites, latitude, longitude, altitude, and AEZs
| No | Region | Latitude (oS) | Longitude (oE) | Altitude | AEZ |
| 1 | Bugesera | 2°09′ | 30°05′ | 1,100–1,780 | Eastern lowlands |
| 2 | Rulindo | 1°44′ | 29°59′ | 1,862–2,438 | Central Highlands |
2.5. Sample determination
Using the (Wonnacott and Wonnacott, 1967) formula, a sample of 381 was obtained as follows:
The next step was determining the number of sample households for each agroecology using the probability proportional to size (PPS) method (Lankamo & Ramalingam, 2025) to equally represent farmer households in each agroecological zone:
Where,
Therefore,
- sample for Bugesera:
- sample size for Rulindo:
Hence, a total of 193 (51%) farmer households were systematically selected from Bugesera dry savannah lowland AEZ, while 188 (49%) were selected from Rulindo highland AEZ.
2.6. Data collection methods
For this study, standardized scheduled interview questions were prepared and entered into Open Data Kit (ODK) software and administered face-to-face using the ODK Collect application to collect data from smallholder households based on the oral consent of the respondents. Precautions were taken to prevent potential bias in the data collection process (Lankamo & Ramalingam, 2025) while standardized measurement tools were prepared based on the literature review. The original questionnaire was also reviewed by two experts in the area (district extensionists), and comments were incorporated. The use of Open Data Kit (ODK) software and its data collection extension, the ODK Collect application, also helped to reduce bias in that the questions were made mandatory to avoid skipping, starting time, and ending time. Finally, other features such as a GPS were included to capture geospatial coordinates, along with socio-economic, bio-physical, and other relevant information. A pre-test survey was conducted with 11 farmer households (3% of the sample farmers) before implementing the final survey, and the feedback was used to refine the questionnaire. Data were collected from farmers for their production season from April to June 2023. Two focus group discussions (FGDs) and key informant interviews (KIIs) (local leaders: 6, opinion leaders: 14, extension officers: 4) were held to collect qualitative data.
2.7. Data analysis methods
We analyzed the data using descriptive statistics and regression. The data collected using ODK were coded and cleaned with R and Stata software. To analyze the practices of adopting the CSAF, descriptive statistics were employed, while an MVP regression model was used to analyze factors associated with smallholder farmers’ adoption decisions for interrelated CSAF practices. Binary logistic regression was utilized to identify the reference category of categorical variables and to analyze the marginal effects of independent variables on each dependent variable.
Farmers adopt various technologies to solve multifaceted agricultural production constraints, so the adoption decision is inherently multivariate. Coincidentally, attempting univariate modeling would be inappropriate. The approach would exclude useful economic information about interdependent and simultaneous adoption decisions (Dorfman, 1996). Therefore, we used a multivariate probit (MVP) model to establish farmers’ choice of interrelated CSAF practices.
Faced with adverse climatic changes, decreasing land productivity, crop failure, and food crises, farmers may adopt a mix of climate-smart farming practices to mitigate climate change and improve productivity rather than relying on a single practice to exploit complementarities among alternatives. Thus, a farmer may opt for various practices rather than one practice. Adoption could result from earlier adopted strategies informing decisions on subsequent practices in the future. In this respect, a model is required to establish the influence of exogenous factors on the adoption of the strategies simultaneously while allowing for the error terms of each of these strategies to be freely correlated. Failure to predict and correct these interrelations in the regression model can lead to biased estimates (Lin et al., 2005; Kassie et al., 2013). Therefore, we employed an MVP model to investigate the interdependence of farmers’ decisions to adopt CSAF.
The MVP model is a generalization of the probit model. It estimates simultaneous correlated binary outcomes (Greene, 2002). A multivariate model is used on more than two outcome variables by adding equations (Donkoh et al., 2019). This study identified seven CSAF practices (home gardens, improved fallow, multipurpose trees, plantation-crop combinations, silvopasture, shelterbelts/windbreaks, and alley cropping). It is anticipated that farmers are more likely to jointly adopt these practices for multiple benefits than to adopt a single practice (Donkoh et al., 2019). Failure to establish the inter-relationships among practices may result in biased estimates of factors correlated with farmers’ adoption decisions (Lin et al., 2005; Kassie et al., 2013).
The MVP is a regression model estimation technique that accounts for the correlation in the error terms by jointly modeling the relationship between a set of covariates with each CSAF practice and estimates a set of binary probit models. The MVP model estimates the relationship between the adoption of CSAF and the potential correlations between unobserved disturbances (Kassie et al., 2009; Ahmed, 2015). Univariate models, such as probit and logistic regression, present drawbacks as they ignore the correlation in the error terms of the adoption estimation equation (Donkoh et al., 2019). These models become inappropriate if adoption practices are interrelated (Belderbos et al., 2004). Failure to account for such unobserved factors and the interdependent relationships of farmers’ adoption choices on CSAF may lead to biased model results (Greene, 2008).
Following Cappellari and Jenkins (2003), an MVP model with seven sets of binary dependent variables (home gardens, improved fallow, multipurpose trees, plantation-crop combinations, silvopasture, shelterbelts/windbreaks, and alley cropping) was estimated as indicated in the formula:
(1)
((k = home gardens (H), improved fallow (F), multipurpose trees (M), plantation-crop combinations (C), silvopasture (P), shelterbelts and windbreaks (S), and alley cropping (A)).
Equation 3 indicates the specification of the binary dependent variable:
(2)
Where, is a latent variable whose counterpart is observed in a binary form measuring the adoption of CSAF practice by household . represents the explanatory variables including demographic characteristics (sex, age, education), asset ownership (radio, mobile phone), institutional variables (training, extension), AEZ (Bugesera, Rulindo), farm characteristics (farm size, livestock) and bio-physical characteristics ((altitude (elevation), distance to rivers)). and are parameters to be estimated, while represents farm characteristics to account for unobserved heterogeneity. represents the multivariate normally distributed stochastic error term (Wooldridge, 2003). In our estimated MVP model, the error terms jointly follow a multivariate normal distribution with zero conditional means (Donkoh et al., 2019). The variance is normalized to the unit, where and the symmetric variance-covariance matrix is specified in the table below:
Table 3. Symmetric table for pairwise correlation of the 7 response variables
Table 3 can be transformed into a generalized form equation (3) as follows:
(3)
is the coefficient of pairwise correlations of the error terms regarding two estimated equations in the MVP regression model. The correlation between the stochastic components of different CSAF practices adopted is represented by the off-diagonal elements in the variance-covariance matrix table 3 (Danso-Abbeam and Baiyegunhi, 2017). The correlation denotes the association between two quantitative variables. It is assumed that the adoption of a new technology depends on another (complementarity or positive correlation) or is influenced by an available set of substitutes (negative correlation) (Khanna, 2001).
2.4. Dependent and explanatory variables
Dependent variables
The dependent variables (CSAFs) we consider include home gardens, improved fallow, multipurpose trees, plantation-crop combinations, silvopasture, shelterbelts/windbreaks, and alley cropping (Table 4).
Table 4. Major CSAF practices
| CSAF practice | Description |
| Home gardens | Refer to multi-storey gardening combining trees with crops in homestead gardens (livestock may or may not be present). |
| Improved fallow | Refers to planting fast-growing species (usually leguminous woody cultivars) during the fallow phase of shifting cultivation. |
| Multipurpose trees | Refer to trees kept and managed for multipurpose use, product, or service in a multiple-output land-use system. |
| Plantation-crop combinations | Integrate crops and trees (such as coconut, cacao, coffee, and fruit trees), shade trees, or fuelwood/fodder trees. |
| Silvopasture | Combines trees and grazing livestock operations on the same land; managed for both forest products and forage, targeting short and long-term economic returns. |
| Shelterbelts/ windbreaks | Refer to strips of trees around farms and gardens planted and managed as part of crop or livestock operations to protect crops, animals, and soil from natural hazards, including wind, excessive rain, sandstorms, droughts, seawater, or floods. |
| Alley cropping | Places trees within agricultural cropland, termed ‘intercropping’, especially in tropical areas, to improve whole-farm yield. |
Source: Adapted from Alavalapati et al. (2001)
Explanatory variables
The independent variables were selected after a thorough literature review of published agricultural innovation adoption. Factors influencing farmers’ choice of any CSAF practice were taken as predictor variables, while CSAF practices are response variables, as described in Table 5, with their expected signs. It should be mentioned that there is a lack of straightforward theory regarding the forward selection of predictors (Nhemachena et al., 2014). However, several previous empirical studies have enabled the nomination of appropriate predictors (Arun and Jun-Ho, 2020). Hence, to discover which factors may be responsible for farmers to adopt CSAF practices, we used the variables summarized in Table 6.
Regional factors such as AEZ and elevation patterns positively or negatively influence the adoption of CSAF, crop, and plant adaptation. Regional biophysical and ecological traits can positively or negatively impact crop and plant growth.
However, understanding how demographic variables like gender, age, civil status, household size, and ubudehe (household poverty level) influence farmers’ adoption choices for CSAF is complex. While male, mature, married, large-sized families and high-income farmers are likely to embrace and effectively practice CSAF, the opposite holds for females, young, single, small-sized families, and low-income farmers. CSAF adoption is vital for reaping higher benefits regarding environmental safety, farm productivity, economic returns, and food sovereignty.
During the research process, the social categorization (“ubudehe”) was counted because it is used to assess and monitor the poverty levels among the Rwandan population. The “ubudehe” process involves categorizing all households into a range of appropriate categories of poverty level, as Table 5 shows (Nizeyimana et al., 2018). In this study, households in categories A and B were classified as high-income earners, while households in categories C, D, and E were classified as low-income earners.
Table 5. “Ubudehe”: Poverty characteristics of households in Rwanda
| Category | Characteristics |
| Category A | This is the category of well-off people. These are households with a household head or spouse who can make diverse life choices from their source of income |
| Category B | This is the category of self-reliant households. These are households with a household head or spouse who has a source of income and can afford household needs for all household members |
| Category C | This is the category of poor households, i.e., households with labor capacity but limited assets to sustain their livelihoods |
| Category D | This category comprises households in extreme poverty, i.e., households that have limited labor capacity and no assets or businesses to sustain their livelihoods |
| Category E | This is a special category that comprises households with no labor capacity due to age, severe disability, or chronic illness, with no other source of income |
Factors such as educational level, farm size, farmer experience in CSAF, possession of communication tools (radio, telephone), livestock size, proximity to rivers, farmer training, and access to extension services play significant roles in predicting positive CSAF adoption outcomes. Highly qualified farmers, those with large-scale farming, and those with accumulated farming experience are more inclined to embrace innovative practices. Possession of communication tools ensures farmers stay informed, livestock husbandry promotes farming diversity, and proximity to rivers aids in hygiene and irrigation during dry spells. Farmer training and access to extension services enhance technology transfer, making farmers well-informed decision-makers capable of accommodating innovations in their activities. Therefore, creating a conducive environment for CSAF adoption involves recognizing regional characteristics and understanding the diverse influences of demographic and socio-economic factors on farmers’ choices. These factors may inform strategy and policy formulation for promoting innovative farming practices and aligning them with farmers’ specific needs and capabilities in different contexts (Table 6).
Table 6. Definition of variables and expected signs for their relationship with the CSAF adoption
| Variables | Description | Expected sign |
| Dependent variables | ||
| Adoption intensity (Y) | Number of CSAF practices adopted (from 1 to 7) | |
| Independent variables | ||
| Location Features | ||
| AEZ | Dummy variable (Bugesera = 1, Rulindo = 0) | +/- |
| Altitude | Continuous variable (number of counts in meters) | +/- |
| Demographic characteristics | ||
| Gender | Dummy variable (male = 1, female = 0) | +/- |
| Age | Continuous variable (age of farmer in years) | +/- |
| Civil status | Dummy variable (married = 1, unmarried = 0) | +/- |
| Education | Dummy variable (literate = 1, illiterate = 0) | + |
| Household size | Discrete variable (number of counts) | +/- |
| “Ubudehe” | Household poverty category (high income =1,
low income = 0) |
+/- |
| Experience | Discrete variable (number of counts in years) | + |
| Farm characteristics | ||
| Farm area | Continuous variable (number of counts in ha) | + |
| Distance from the river | Dummy variable (˂ 500m, ˃ 500m) | + |
| Asset ownership | ||
| Radio | Dummy variable (radio = 1, no = 0) | + |
| Mobile phone | Dummy variable (mobile phone = 1, no = 0) | + |
| Livestock | Discrete variable (number of counts) | + |
| Institutional variables | ||
| Training | Dummy variable (yes = 1, no = 0) | + |
| HH Ext. visit | Dummy variable (yes = 1, no = 0) | + |
- Results
3.1. Descriptive statistics and correlation results
Table 7 shows the average differences between farmers who practice CSAF (adopters) and those who do not practice CSAF (non-adopters) in the study areas. On average, adopters had a larger size of livestock (≈ 2 cows) compared to non-adopters (≈ 1 cow) and received training on CSAF. The average age of adopters was 44 years, while that of non-adopters was 43 years. There was a significant difference in livestock ownership between CSAF adopters and non-adopters (p ≤ 0.1). There was also a variation between CSAF adopters and non-adopters regarding training (p ≤ 0.05) in modern, smart farming; other variables failed to show any significant differences between CSAF adopters and non-adopters in study areas.
Table 7. Descriptive statistics and mean difference between CSAF adopters and non-adopters
| Variables | Adopters | Non-adopters | Variation | P>|t| |
| AEZ (1/0) | 0.51(0.50) | 0.50(0.50) | 0.01 | 0.77 |
| Altitude (m) | 1640.47(239.28) | 1640.05(219.68) | 0.42 | 0.98 |
| Gender (1/0) | 0. 67 (0.50) | 0.66(0.98) | 0.01 | 0.88 |
| Age (years) | 44.19(15.05) | 42.85(14.67) | 1.34 | 0.38 |
| Civil status (1/0) | 0.70(0.45) | 0.77(0.42) | -0.07 | 0.15 |
| Education (1/0) | 0.56(0.49) | 0.57(0.49) | -0.01 | 0.88 |
| Household size (counts) | 3.99(1.86) | 4.09(1.86) | -0.10 | 0.58 |
| “Ubudehe” (1/0) | 0.58(0.49) | 0.54(0.50) | 0.04 | 0.45 |
| Farm area (ha) | 1.15(1.62) | 1.04(1.21) | 0.11 | 0.46 |
| Experience (years) | 2.48(3.13) | 2.50(3.42) | -0.02 | 0.96 |
| Radio (1/0) | 0.50(0.50) | 0.48(0.50) | 0.02 | 0.67 |
| Mobile phone (1/0) | 0.73(0.44) | 0.71(0.45) | 0.02 | 0.66 |
| Livestock (counts) | 1.81(6.32) | 0.98(1.70) | 0.83** | 0.06 |
| Distance from river (<500m/>500m) | 0.12(0.33) | 0.14(0.35) | -0.02 | 0.59 |
| Training (1/0) | 0.18(0.39) | 0.11(0.31) | 0.07* | 0.04 |
| HH Ext. Visit (1/0) | 0.24(0.43) | 0.18(0.39) | 0.06 | 0.16 |
*significant at 0.05, **significant at 0.1.
Bold values emphasize the statistically significant values.
Farmers were post-stratified into four categories as per the total landholding of these farmers (Table 8). The marginal farmers had an average landholding of 0.13ha, smallholder farmers had an average landholding of 0.67ha, medium farmers had an average landholding of 1.09ha, and the large farmers had an average landholding of 3.42ha. Overall, the average landholding is estimated to be 1.09ha. Alternatively, Bugesera had a larger average of landholding than Rulindo, with 1.48ha and 0.69ha, respectively.
Table 8. Categorization of surveyed farmers
| Farmer’s category | Bugesera | Rulindo | Total | |||
| Freq. (%) | Average holding size (ha) | Freq. (%) | Average holding size (ha) | Freq. (%) | Average holding size (ha) | |
| Marginal (< 0.5ha) | 41(21.24) | 0.13 | 104(55.32) | 0.14 | 145(38.06) | 0.13 |
| Small (0.5-<1ha) | 60(31.09) | 0.63 | 38(20.21) | 0.74 | 98(25.72) | 0.67 |
| Medium (1-2ha) | 39(20.21) | 1.08 | 22(11.70) | 1.10 | 61(16.01) | 1.09 |
| Large (>2ha) | 53(27.46) | 3.79 | 24(12.77) | 2.60 | 77(20.21) | 3.42 |
| Total | 193(100.00) | 1.48 | 188(100.00) | 0.69 | 381(100.00) | 1.09 |
Table 9 presents the distribution of CSAF practices among adopters and non-adopters in the study areas, grouped by AEZ, portraying the dominance and local variations in practicing CSAF. Regional categorization of practitioners enables an understanding of the CSAF practices in the study areas. The table displays the number of adopters engaged in different CSAF practices in the study areas (Bugesera and Rulindo). In Bugesera, most adopters recorded practicing home gardens (72 adopters), while in Rulindo, most farmers practiced shelterbelts or windbreaks (72 adopters). A smaller number of adopters practiced alley cropping in Bugesera (2 adopters), while in Rulindo, a smaller number practiced improved fallow and silvopasture (2 adopters), respectively. These low numbers are explained by agroecological characteristics, notably improved fallow and silvopasture, which are unfavorable on the steep slopes and highly fragmented land in the Rulindo highland. In the Bugesera drought-prone zone, alley cropping with tree hedgerows is less adopted due to farmers’ unawareness of the potential of CSAF to increase the whole-farm yield. All the CSAF adoptions in study areas showed statistically significant differences except the plantation-crop-combination practice (p = 0.949).
Table 9. Distribution of CSAF adopters and non-adopters in the study areas
| Practice | Bugesera | Rulindo | p-value | ||
| Adopters | Non-adopters | Adopters | Non-adopters | ||
| H. gardens | 72 | 121 | 49 | 139 | 0.018 |
| A. cropping | 2 | 191 | 36 | 152 | 0.000 |
| I. fallow | 56 | 137 | 2 | 186 | 0.000 |
| M. trees | 34 | 159 | 13 | 175 | 0.001 |
| P. combination | 11 | 182 | 11 | 177 | 0.949 |
| Shelterbelts | 23 | 170 | 72 | 116 | 0.000 |
| Silvopasture | 8 | 185 | 2 | 186 | 0.060 |
Table 10 shows the relationship between the seven CSAF practices (pairwise correlations). All the pairwise coefficients in the matrix were positively correlated, indicating complementarity among the seven innovative farming practices. The highest positive correlation was between multipurpose trees and silvopasture (rho = 99%), while the lowest was between the plantation-crop-combination and alley cropping, and the plantation-crop-combination and improved fallow (rho = 65%), respectively.
Table 10. Correlation matrix of CSAF from the MVP model
| H. gardens | A. cropping | I. fallow | M. trees | P. combination | Shelterbelts | Silvopasture | |
| H. gardens | 1.000 | ||||||
| A. cropping | 0.815 | 1.000 | |||||
| I. fallow | 0.783 | 0.906 | 1.000 | ||||
| M. trees | 0.917 | 0.895 | 0.964 | 1.000 | |||
| P. combination | 0.971 | 0.652 | 0.650 | 0.829 | 1.000 | ||
| Shelterbelts | 0.876 | 0.907 | 0.696 | 0.782 | 0.771 | 1.000 | |
| Silvopasture | 0.857 | 0.909 | 0.991 | 0.990 | 0.745 | 0.745 | 1.000 |
3.2. Regression results
Different factors can influence farmers’ decisions to adopt a new farming technology. The MVP model (Table 11) exhibits the regression results justifying the variables determining farmers’ adoption choices for CSAF. The model included seven dependent variables (CSAF practices). Sixteen model factors (regressors) were included in the model, of which three were discrete, ten were dummy, and three were continuous (Table 11).
The agroecological zone (AEZ) variable showed a significant positive relationship with farmers’ choices to adopt alley cropping (p < 0.05) and a statistically significant but negative relationship with improved fallow (p < 0.01) adoption. This implies that farmers in the Bugesera (eastern lowland) region tended to adopt this CSAF practice more than their counterparts in the Rulindo (central highland) region. Inversely, results suggest that farmers in the Rulindo region tended to adopt improved fallow more than farmers in the Bugesera region. Disseminating innovative farming technologies to farmers is critical in propelling and diversifying productivity through CSAF practices.
Landscape elevation (altitude) showed a positive and significant relationship with farmers’ choices to adopt shelterbelts or windbreaks (p < 0.10) but was insignificant in explaining the adoption of the other CSAF practices. They were more practiced in Rulindo (central highland) than in Bugesera (eastern lowland). This could be attributed to the fact that during storms, strong winds in high altitudes damage crops (maize, sorghum, wheat, banana plantations) and blow off the rooftops of substandard residential houses and public structures (schools, health centers).
The farmer’s gender significantly influenced CSAF adoption choices, with a negative impact on multipurpose tree adoption (p < 0.10) and a positive influence on silvopasture adoption (p < 0.10). Female farmers were most likely to adopt multipurpose trees, possibly due to their central role as household caretakers in SSA — multipurpose tree farming aimed to provide various household products and services (utilities), aligning with women’s responsibilities for family survival and well-being. Conversely, male farmers were mostly inclined to adopt silvopasture, reflecting their predominant role as decision-makers in cattle keeping and overseeing large-scale farming activities. This gender-based divergence in adoption preferences underscores the distinct roles and responsibilities within households, shaping farmers’ choices in CSAF practices.
The farmer’s age significantly correlated with the adoption of CSAF practices: negatively for improved fallow and multipurpose tree farming (p < 0.10) and positively for plantation-crop combinations (p < 0.10). Younger farmers were more inclined to adopt improved fallow, engage in off-farm activities, and allow longer fallow periods. This trend is evident in Bugesera, where large unexploited farms are owned by youth residing in urban areas. Similarly, younger farmers favored CSAF multipurpose tree practices, seeking monetary income from diverse products. Conversely, older farmers adopt plantation-crop combinations to maximize utility from combined food crops and tree products. This suggests age-related preferences in adopting CSAF practices, aligning with income-generation priorities for the youth and utility maximization for older farmers.
There was an inverse relationship between farm size and home garden (p < 0.10) practices and a positive relationship with the adoption of multipurpose trees (p< 0.05) and shelterbelts/windbreaks (p < 0.05) practices. This implies that farmers with larger land sizes had a lower probability of adopting home garden trees than farmers with smaller farms. This can be attributed to the fact that farmers with large land areas tend to transform subsistence farming into a business. Hence, harvests obtained in commercial farming outweigh those received from home gardening. In this regard, farmers with large farm sizes combine and diversify tree species within crops and on-farm boundaries to serve multiple purposes.
Farmer farming experience was negative and significantly associated with home garden (p < 0.05) and improved fallow (p < 0.01), but positively and significantly associated with plant-crop-combination and insignificant in explaining the adoption of other CSAF practices. This result implies that home garden farming requires less expertise and is practiced in small spaces in homesteads. Similarly, improved fallows imply that farmlands are left uncultivated for a period to restore soil fertility. On the other hand, a plantation-crop combination requires knowledge, expertise, and experience accumulated over several years in farming.
Radio had a negative and significant relationship with the home garden (p < 0.05) and was insignificant in explaining the adoption of other CSAF practices. This finding did not meet the a priori expectation since radio and other communication channels are meant to communicate information to the audience and influence technology uptake by farmers. Usually, communication channels inform farmers of the latest technologies and weather forecasts by enabling them to make informed decisions about planting, irrigation, pest control, and harvesting, and to prevent crop failure due to adverse weather events (drought and floods) and pest infestation.
Livestock had a negative and significant relationship with multipurpose trees (p < 0.10) and was insignificant in explaining the adoption of other CSAF practices. This finding is explained by the fact that large-scale cattle keeping is practiced in rangeland (pastures), whereas a small number of cows are kept in cowsheds (homesteads), fed with fodder from multipurpose trees and crop residues.
Farm-river distance had an inverse relationship with home garden practices (p < 0.10) and was insignificant in explaining the adoption of other CSAF practices. This result implies that proximity to rivers for irrigation in drought-prone regions is important for sustainable farming. Inversely, using household water, home gardens require watering in the morning and evening. In this respect, long distances to rivers may not directly affect home gardens.
Table 11: Regression results for the determinants of CSAF adoption
| Covariate | Home garden | Alley cropping | Improved fallow | Multipurpose trees | Plant crop combination | Shelterbelts (windbreaks) | Silvopasture | |||||||
| Coef. | SE | Coef. | SE | Coef. | SE | Coef. | SE | Coef. | SE | Coef. | SE | Coef. | SE | |
| AEZ | (0.138) | 0.122 | 0.195b | 0.076 | (0.380)a | 0.380 | 0.019 | 0.086 | 10-4 | 0.062 | 0.134 | 0.109 | (0.006) | 0.043 |
| Altitude | (10-4) | 10-4 | (10-4) | 10-4 | 10-4 | 10-4 | (10-4) | 10-4 | (10-4) | 10-4 | 10-4c | 10-4 | (10-4) | 10-4 |
| Gender | (0.006) | 0.052 | 0.008 | 0.032 | (0.045) | 0.037 | (0.067)c | 0.037 | 0.020 | 0.026 | 0.064 | 0.047 | 0.031c | 0.018 |
| Age | 10-4 | 0.002 | 10-4 | 0.001 | (0.002)c | 0.001 | (0.002)c | 0.001 | 0.002c | 0.001 | 10-4 | 0.002 | 0.001 | 0.001 |
| M. (status) | (0.014) | 0.068 | (0.048) | 0.042 | (0.044) | 0.048 | 0.047 | 0.048 | (0.009) | 0.034 | (0.005) | 0.061 | 0.010 | 0.024 |
| Education | (0.024) | 0.062 | (0.027) | 0.039 | 0.031 | 0.045 | 0.035 | 0.044 | (0.017) | 0.032 | 0.016 | 0.056 | 0.031 | 0.022 |
| HH size | (0.023) | 0.016 | 0.010 | 0.010 | (0.002) | 0.011 | 0.010 | 0.011 | 0.004 | 0.008 | (0.003) | 0.014 | (0.006) | 0.006 |
| “Ubudehe” | 0.082 | 0.057 | (0.058) | 0.035 | (0.007) | 0.041 | (0.032) | 0.040 | (0.013) | 0.029 | 0.060 | 0.051 | (0.014) | 0.020 |
| Farm size | (0.043)c | 0.022 | (0.017) | 0.014 | (0.014) | 0.016 | 0.031b | 0.016 | (0.018) | 0.011 | 0.047b | 0.020 | 0.009 | 0.008 |
| Experience | (0.017)b | 0.007 | 0.004 | 0.005 | (0.014)a | 0.005 | 0.008 | 0.005 | 0.013a | 0.004 | 0.002 | 0.007 | (0.002) | 0.003 |
| Radio | (0.126)b | 0.055 | 0.037 | 0.034 | 0.042 | 0.039 | 0.014 | 0.038 | (0.005) | 0.028 | 0.009 | 0.049 | (0.010) | 0.019 |
| M. (phone) | 0.015 | 0.062 | 0.056 | 0.038 | 0.014 | 0.044 | (0.035) | 0.043 | (0.026) | 0.031 | 0.011 | 0.055 | (0.003) | 0.022 |
| Livestock | 0.003 | 0.006 | 0.003 | 0.004 | 0.007 | 0.004 | (0.008)c | 0.004 | 0.002 | 0.003 | (0.007) | 0.006 | 10-4 | 0.002 |
| D.(rivers) | (0.141)c | 0.074 | (0.046) | 0.046 | 0.052 | 0.053 | 0.031 | 0.052 | (0.017) | 0.037 | 0.103 | 0.066 | (0.001) | 0.026 |
| Training | 0.074 | 0.093 | (0.036) | 0.058 | 0.005 | 0.067 | (0.018) | 0.066 | (0.046) | 0.047 | 0.009 | 0.084 | (0.001) | 0.033 |
| Extension | (0.048) | 0.079 | 0.011 | 0.049 | (0.038) | 0.056 | 0.061 | 0.055 | 0.037 | 0.040 | 0.004 | 0.071 | 0.023 | 0.027 |
| Constant | 1.019a | 0.362 | 0.201 | 0.226 | 0.021 | 0.260 | 0.317 | 0.256 | 0.068 | 0.183 | (0.732)b | 0.326 | 0.011 | 0.127 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
3.3. Marginal effects
The MVP parameter estimates the likelihood of the outcome occurrence (Arun and Jun-Ho, 2020). This likelihood of the occurrence of the outcome cannot be quantified, but the marginal effect can be computed for each model, which produces the quantified influence of the variables (ibid.).
3.3.1. Marginal effect of the change in the home garden practices model
Table 12 portrays the marginal effects of predictor variables for the home garden practices model. Possession of a radio as a tool source of information and the shorter farm-river distance increased the probability of home garden practices’ adoption, whereas farm size and experience decreased it. The largest effect was found for the farm-river distance, while possessing a radio was ranked second. Thus, a unit increase in farming near rivers would raise the probability of adopting home garden practices by about 14%. Similarly, a unit increase in radio possession would raise the likelihood of adoption (12.5%). Inversely, CSAF farming experience and farm size have negative marginal effects on the probability of home garden adoption by about 1.6% and 4.2%, respectively.
Table 12. Marginal effect after MVP analysis due to the change in home garden practices
| Variable | dy/dx | SE | t | P>|t| | 95% CI |
| AEZ | 0.138 | 0.121 | 1.14 | 0.256 | -0.100-0.377 |
| Altitude | -0.000 | 0.000 | -0.67 | 0.504 | -0.000-0.000 |
| Gender | 0.005 | 0.051 | 0.12 | 0.908 | -0.095-0.107 |
| Age | 0.000 | 0.001 | 0.11 | 0.914 | -0.003-0.003 |
| M. (status) | 0.013 | 0.067 | 0.20 | 0.841 | -0.119-0.146 |
| Education | 0.023 | 0.062 | 0.38 | 0.705 | -0.099-0.146 |
| HH size | -0.023 | 0.015 | -1.45 | 0.148 | -0.054-0.008 |
| Ubudehe | -0.082 | 0.056 | -1.44 | 0.150 | -0.193-0.029 |
| Farm size | -0.042c | 0.022 | -1.91 | 0.056 | -0.086-0.001 |
| Experience | -0.016b | 0.007 | -2.24 | 0.026 | -0.031–0.001 |
| Radio | 0.125b | 0.054 | 2.30 | 0.022 | 0.018-0.233 |
| M. (phone) | -0.014 | 0.061 | -0.24 | 0.809 | -0.136-0.106 |
| Livestock | 0.002 | 0.006 | 0.43 | 0.670 | -0.009-0.015 |
| D.(rivers) | 0.140c | 0.073 | 1.91 | 0.056 | -0.003-0.285 |
| Training | -0.002 | 0.093 | -0.03 | 0.979 | -0.185-0.180 |
| Extension | 0.048 | 0.0785 | 0.62 | 0.539 | -0.106-0.202 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
3.3.2. Marginal effect for alley cropping practices model
Ceteris paribus, the probability of adopting alley cropping in Bugesera would be lower by about 19.4% compared to a typical farmer household in Rulindo, the comparison group (Table 13).
Table 13. Marginal effect after MVP analysis due to the change in alley cropping practices
| Variable | dy/dx | SE | t | P>|t| | 95% CI |
| AEZ | -0.194b | 0.075 | -2.56 | 0.011 | -0.343–0.045 |
| Altitude | -0.000 | 0.000 | -0.67 | 0.503 | -0.000-0.000 |
| Gender | -0.007 | 0.032 | -0.24 | 0.813 | -0.071-0.055 |
| Age | 0.000 | 0.001 | 0.16 | 0.870 | -0.002-0.002 |
| M. (status) | 0.047 | 0.042 | 1.14 | 0.256 | -0.035-0.130 |
| Education | 0.026 | 0.038 | 0.69 | 0.491 | -0.049-0.103 |
| HH size | 0.010 | 0.009 | 1.03 | 0.306 | -0.009-0.029 |
| Ubudehe | 0.057 | 0.035 | 1.62 | 0.106 | -0.012-0.127 |
| Farm size | -0.017 | 0.013 | -1.24 | 0.216 | -0.044-0.010 |
| Experience | 0.004 | 0.004 | 0.87 | 0.384 | -0.005-0.013 |
| Radio | -0.036 | 0.034 | -1.08 | 0.281 | -0.103-0.030 |
| M. (phone) | -0.055 | 0.038 | -1.45 | 0.149 | -0.131-0.019 |
| Livestock | 0.002 | 0.003 | 0.67 | 0.503 | -0.005-0.010 |
| D.(rivers) | 0.000 | 0.045 | 1.01 | 0.313 | -0.043-0.136 |
| Training | 0.036 | 0.058 | 0.62 | 0.535 | -0.078-0.150 |
| Extension | -0.011 | 0.049 | -0.23 | 0.817 | -0.107-0.085 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
3.3.3. Marginal effect for the improved fallow practices model
The result of the marginal analysis of the improved fallow practices model is presented in Table 14. Except for the agroecological zone variable, the age and experience of the farmers imparted negative marginal effects. All else equal, the likelihood of adopting improved fallow in Bugesera would be about 37.9% higher than in a typical farmer household in Rulindo, the comparison group. A unit increase in farmers’ age and CSAF farming experience would decrease the likelihood of adopting improved fallow by about 0.2% and 1.4%, respectively.
Table 14. Marginal effect after MVP analysis due to the change in improved fallow practices
| Variable | dy/dx | SE | t | P>|t| | 95% CI |
| AEZ | 0.379a | 0.087 | 4.35 | 0.000 | 0.208-0.551 |
| Altitude | 0.000 | 0.000 | 1.55 | 0.122 | -0.000-0.000 |
| Gender | 0.045 | 0.037 | 1.23 | 0.221 | -0.027-0.118 |
| Age | -0.002c | 0.001 | -1.85 | 0.065 | -0.004-0.000 |
| M. (status) | 0.043 | 0.048 | 0.90 | 0.368 | -0.051-0.139 |
| Education | -0.030 | 0.044 | -0.69 | 0.493 | -0.118-0.057 |
| HH size | -0.001 | 0.011 | -0.16 | 0.870 | -0.024-0.020 |
| Ubudehe | 0.007 | 0.040 | 0.17 | 0.862 | -0.073-0.087 |
| Farm size | -0.013 | 0.016 | -0.86 | 0.393 | -0.045-0.017 |
| Experience | -0.014a | 0.005 | -2.65 | 0.009 | -0.024–0.003 |
| Radio | -0.042 | 0.039 | -1.08 | 0.282 | -0.119-0.034 |
| M. (phone) | -0.014 | 0.044 | -0.33 | 0.744 | -0.101-0.072 |
| Livestock | 0.006 | 0.004 | 1.48 | 0.139 | -0.002-0.015 |
| D.(rivers) | 0-.052 | 0.052 | -1.00 | 0.320 | -0.156-0.051 |
| Training | -0.004 | 0.066 | -0.07 | 0.943 | -0.136-0.126 |
| Extension | 0.038 | 0.056 | 0.68 | 0.500 | -0.072-0.148 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
3.3.4. Marginal effect for the multipurpose tree practices model
Gender and farm size increased the probability of multipurpose tree practices; in contrast, the age of the farmers and livestock size decreased it (Table 15). Moreover, with the farmer being male, the probability of adoption increased by 6.6% compared to a female farmer. A unit increase in farm size would increase the likelihood of adopting multipurpose tree farming by about 3.1%. In contrast, ceteris paribus, a unit increase in the farmer’s age and livestock would decrease the chances of adopting multipurpose trees by about 0.2% and 0.8%, respectively.
Table 15. Marginal effect after MVP analysis due to the change in multipurpose tree practices
| Variable | dy/dx | SE | t | P>|t| | 95% CI |
| AEZ | -0.019 | 0.085 | -0.22 | 0.823 | -0.187-0.149 |
| Altitude | -0.000 | 0.000 | -0.84 | 0.401 | -0.000-0.000 |
| Gender | 0.066c | 0.036 | 1.82 | 0.069 | -0.005-0.138 |
| Age | -0.002c | 0.001 | -1.79 | 0.075 | -0.004-0.000 |
| M. (status) | -0.046 | 0.047 | -0.98 | 0.326 | -0.140-0.046 |
| Education | -0.035 | 0.044 | -0.80 | 0.427 | -0.121-0.051 |
| HH size | 0.010 | 0.011 | 0.90 | 0.367 | -0.011-0.032 |
| Ubudehe | 0.032 | 0.040 | 0.80 | 0.421 | -0.046-0.111 |
| Farm size | 0.031b | 0.015 | 1.97 | 0.050 | 0.0000-0.062 |
| Experience | 0.008 | 0.005 | 1.58 | 0.116 | -0.002-0.018 |
| Radio | -0.013 | 0.038 | -0.36 | 0.717 | -0.089-0.061 |
| M. (phone) | 0.034 | 0.043 | 0.80 | 0.422 | -0.050-0.120 |
| Livestock | -0.008c | 0.004 | -1.91 | 0.056 | -0.017-0.000 |
| D.(rivers) | -0.031 | 0.051 | -0.60 | 0.547 | -0.133-.070 |
| Training | 0.018 | 0.065 | 0.28 | 0.780 | -0.110-0.147 |
| Extension | -0.060 | 0.055 | -1.10 | 0.272 | -0.169-0.048 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
3.3.5. Marginal effect for plantation-crop-combination practices model
The results of marginal effects for plantation-crop-combination practices are presented in Table 16. Both variables (age and experience) positively influence plantation-crop combination practices. If a farmer’s experience in CSAF increased by a unit, the probability of adoption equally increased by about 1.2%. Similarly, with an increase in a farmer’s age by a unit, the likelihood of adopting plantation-crop-combination practices would be increased by about 0.1%.
Table 16. Marginal effect after MVP analysis due to the change in plant crop combination practices
| Variable | dy/dx | SE | t | P>|t| | 95% CI |
| AEZ | -0.000 | 0.061 | -0.00 | 0.997 | -0.121-0.120 |
| Altitude | -0.000 | 0.000 | -0.35 | 0.725 | -0.000-0.000 |
| Gender | -0.020 | 0.026 | -0.77 | 0.441 | -0.071-0.031 |
| Age | 0.001c | 0.000 | 1.82 | 0.069 | -0.000-0.003 |
| M. (status) | 0.009 | 0.034 | 0.27 | 0.785 | -0.057-0.076 |
| Education | 0.017 | 0.031 | 0.55 | 0.585 | -0.044-0.079 |
| HH size | 0.004 | 0.008 | 0.51 | 0.609 | -0.011-0.019 |
| Ubudehe | 0.013 | 0.028 | 0.46 | 0.648 | -0.043-0.069 |
| Farm size | -0.017 | 0.011 | -1.58 | 0.116 | -0.040-0.004 |
| Experience | 0.012a | 0.003 | 3.42 | 0.001 | 0.005-0.020 |
| Radio | 0.004 | 0.027 | 0.18 | 0.858 | -0.049-0.059 |
| M. (phone) | 0.025 | 0.031 | 0.82 | 0.413 | -0.035-0.086 |
| Livestock | 0.001 | 0.003 | 0.55 | 0.580 | -0.004-0.007 |
| D.(rivers) | 0.016 | 0.037 | 0.45 | 0.653 | -0.056-0.089 |
| Training | 0.045 | 0.047 | 0.97 | 0.332 | -0.046-0.138 |
| Extension | -0.036 | 0.039 | -0.93 | 0.354 | -0.115-0.041 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
3.3.6. Marginal effect for shelterbelts/windbreaks practices model
For every additional hectare of land, the probability of adopting shelterbelts/windbreak practices increased by about 4.7% (Table 17). For altitude, any increase doesn’t increase or decrease the adoption of shelterbelts/windbreak practices (0%).
Table 17. Marginal effect after MVP analysis due to the change in shelterbelt/windbreak practices
| Variable | dy/dx | SE | t | P>|t| | 95% CI |
| AEZ | -0.133 | 0.109 | -1.22 | 0.223 | -0.348-0.081 |
| Altitude | 0.000c | 0.000 | 1.95 | 0.052 | -0.000-0.000 |
| Gender | -0.063 | 0.046 | -1.37 | 0.171 | -0.155-0.027 |
| Age | 0.000 | 0.001 | 0.26 | 0.797 | -0.002-0.003 |
| M. (status) | 0.004 | 0.060 | 0.08 | 0.938 | -0.114-0.124 |
| Education | -0.016 | 0.056 | -0.29 | 0.771 | -0.126-0.094 |
| HH size | -0.002 | -0.002 | -0.18 | 0.855 | -0.030-0.025 |
| Ubudehe | -0.059 | 0.051 | -1.17 | 0.244 | -0.160-0.040 |
| Farm size | 0.047b | 0.020 | 2.35 | 0.019 | 0.007-0.086 |
| Experience | 0.002 | 0.006 | 0.37 | 0.714 | -0.010-0.015 |
| Radio | -0.009 | 0.049 | -0.19 | 0.849 | -0.105-0.087 |
| M. (phone) | -0.011 | 0.055 | -0.20 | 0.842 | -0.120-0.097 |
| Livestock | -0.006 | 0.005 | -1.17 | 0.241 | -0.017-0.004 |
| D.(rivers) | -0.102 | 0.066 | -1.55 | 0.121 | -0.233-0.027 |
| Training | -0.009 | 0.083 | -0.11 | 0.911 | -0.174-0.155 |
| Extension | -0.004 | 0.070 | -0.06 | 0.952 | -0.143-0.134 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
3.3.7. Marginal effect for the silvopasture practices model
Table 18 illustrates that, ceteris paribus, the chances of adopting silvopasture practices would be lower by about 3.1% for a male farmer compared to a female farmer in the comparison group. This could be attributed to women’s empowerment programs to reduce gender inequalities and provide grants to make them entrepreneurs and decision-makers in their households and agents of change in their communities.
Table 18. Marginal effect after MVP analysis due to the change in silvopasture practices
| Variable | dy/dx | SE | t | P>|t| | 95% CI |
| AEZ | 0.005 | 0.042 | 0.13 | 0.896 | -0.078-0.089 |
| Altitude | -0.000 | 0.000 | -0.13 | 0.896 | -0.000-0.000 |
| Gender | -0.031c | 0.018 | -1.73 | 0.084 | -0.066-0.004 |
| Age | 0.000 | 0.000 | 0.83 | 0.406 | -0.000-0.001 |
| M. (status) | -0.009 | 0.023 | -0.42 | 0.673 | -0.056-0.036 |
| Education | -0.031 | 0.021 | -1.43 | 0.154 | -0.074-0.011 |
| HH size | -0.005 | 0.005 | -0.99 | 0.321 | -0.016-0.005 |
| Ubudehe | 0.013 | 0.019 | 0.69 | 0.493 | -0.025-0.052 |
| Farm size | 0.009 | 0.007 | 1.21 | 0.227 | -0.005-0.024 |
| Experience | -0.001 | 0.002 | -0.77 | 0.441 | -0.007-0.003 |
| Radio | 0.010 | 0.019 | 0.53 | 0.598 | -0.027-0.047 |
| M. (phone) | 0.002 | 0.021 | 0.12 | 0.903 | -0.039-0.044 |
| Livestock | 0.000 | 0.002 | 0.07 | 0.948 | -0.004-0.004 |
| D.(rivers) | 0.000 | 0.025 | 0.04 | 0.970 | -0.049-0.051 |
| Training | 0.001 | 0.032 | 0.04 | 0.971 | -0.062-0.065 |
| Extension | -0.022 | 0.027 | -0.82 | 0.412 | -0.076-0.031 |
SE = standard error
Note: a significant at 0.01, b significant at 0.05, and c significant at 0.1
- Discussion
The study reveals that AEZ is positively associated with alley-cropping adoption but negatively linked to improved fallow adoption. Alley cropping, placing trees within crops, is favored in the lowland zones of Bugesera compared to the central highland zone of Rulindo. This aligns with observations that alley cropping thrives in humid lowlands (Tengnas, 1994). Conversely, improved fallow is less common in Rulindo (due to fragmented highlands due to population pressure) but is largely practiced in Bugesera (characterized by larger farms). This coincides with the notion that improved fallow is suited to relatively large farms (Quinion et al., 2010). The adoption of new technologies is thus driven by landholding size, which underlines the benefit gains associated with larger landholdings.
The study results showed that altitude was positively and significantly associated with farmers’ choices to adopt shelterbelts/windbreak CSAF practices. In Bugesera, shelterbelts/windbreaks are common practice due to its large landholding system and many village settlements without trees around homesteads and schools to prevent extreme heat, wind erosion, solar radiation, and violent storms, which render families homeless by blowing off the rooftops of residential houses, and public edifices (health centers, and schools). (Tengnas, 1994) found comparable results that large windbreaks are most suited in large-scale farming areas.
The gender of the farmer was negative and significantly influenced farmers’ multipurpose tree adoption choices, but positively influenced silvopasture adoption. This signifies that male farmers were less likely to adopt multipurpose trees and more likely to adopt silvopasture. (Glover et al., 2013) emphasized that, if not given due consideration in advance, the different roles played by men and women in a community can hinder the adoption of CSAF. These findings align with (Tengnas, 1994), who reported that Kenyan women focus on and show interest in fruit trees and trees that yield fuelwood (multipurpose tree farming). Contrariwise, since construction is mostly in the man’s domain, he naturally shows a strong interest in trees yielding good poles and timber. Idem for livestock keeping, which is considered a man’s business.
The age of farmers can sometimes act as a hindrance to adopting a recommended novel technology (Mbata, 2001). The results showed that the farmer’s age is significant, negatively related to the adoption of improved fallow and multipurpose tree farming, and positively associated with the adoption of plantation-crop-combination. It signifies that the older the farmers, the more likely they are to adopt a plant-crop-combination. In contrast, the negative coefficients of age for improved fallow and multipurpose tree practices might be attributed to the fact that youth are relatively innovative and easily adapt to change compared to older generations (Piya et al., 2013; Ali and Erenstein, 2017), and the positive relationship between age and plantation-crop-combination practices might be associated with experience in CSAF, cumulative knowledge, know-how and skills (Ahmed, 2016). Additionally, as noted in various other research works, the influence of age in the decision to adopt a new farming practice was both positive (Nhemachena et al., 2014; Ahmed, 2016; Mulwa et al., 2017) and negative (Mbata, 2001; Piya et al., 2013; Tesfaye and Seifu, 2016; Ali and Erenstein, 2017).
Home garden adoption is negatively associated with farm size, while multipurpose trees and shelterbelts/windbreaks show a positive association. Home gardens are more appealing in regions where most people live and rely on rainfed subsistence farming, and subsistence food needs are unmet by existing systems and are commonly practiced around homesteads with stable settlements (Kirby and Potvin, 2007). These gardens yield fruits and provide shade and ornamental trees. Larger farm sizes, characterized by labor shortages to exploit the land, prompt farmers to adopt CSAF practices like multipurpose trees or shelterbelts/windbreaks (Glover et al., 2013). In treeless landscapes, strong winds damage crops and houses, making shelterbelts essential for protection and microclimate improvement. Trees around homesteads, hospitals, and schools safeguard buildings from winds, enhance aesthetics, and serve functional roles, such as animal protection, keeping highways safer, and beautifying landscapes and boundary demarcation (Tengnas, 1994).
Farmer farming experience was negative and significantly associated with home gardens and improved fallow, but positively and significantly associated with the plantation-crop combination. These findings suggest that the accumulation of experiences possessed by the farmer determines the CSAF adoption choice. (Tega and Bojago, 2023) found that the farmers’ experiences affected home gardens and improved fallow and plantation-crop-combination of CSAF with differing coefficient signs. Farmers knowledgeable with CSAF practices and understanding the benefits of CSAF are more likely to adopt plantation-crop-combination practices than home gardens and improved fallow. Comparable results by Mignouna et al. (2011) and Kariyasa et al. (2013) found that farmers become conversant with evaluating technical information over time as they gain and accumulate know-how via practice.
Unexpectedly, radio shows a negative and significant relationship with home gardens, possibly due to ineffective programs on home garden practices and lower farmer dependency on this technology for food security. Radios broadcast weather forecasts for conventional agriculture, urging and mobilizing farmers for planting seasons, irrigation, harvesting, and market opportunities. Comparable results were reported in Indonesia (Paramita et al., 2014) that radio, mobile phones, and television are top choices for information, although farmers prefer practical farm demonstrations. For enhanced adoption of CSAF practices, information dissemination through extension agents and media like broadcasters connects farmers with transformative agricultural solutions. These findings align with studies emphasizing radio’s power as an effective tool for promoting agriculture and rural transformation, especially in developing countries (Nakabugu, 2001; Chapman et al., 2003; FAO, 2011).
Livestock exhibits a negative and significant association with multipurpose tree practices. This result implies that animals may be kept in cowsheds and fed with fodder from multipurpose trees instead of traditional grazing. Trees are feed sources when the grass is scarce (zero-grazing areas), especially for cowshed-kept animals (Tengnas, 1994). Inversely, (Tega and Bojago, 2023) found a positive association between livestock and CSAF practice adoption, suggesting that cattle herds represent household wealth in pastoral farming communities, prompting investment in CSAF. Studies by Nkamleu and Manyong (2005), Jara-Rojas et al. (2020), and Bishaw et al. (2022) support this result, indicating that a larger livestock herd size ownership increases the demand for feeds and wood for livestock infrastructure (housing, manger, water trough, calf pens, feed store room), met by multipurpose trees. In low and poor-resource countries (with rapid population growth and food insecurity), relying on rainfed subsistence farming and complementing grazing practices with trees as a forage source is helpful and crucial. Patel et al. (2024) argued that silvopasture systems that deploy multipurpose trees and grass species are considered suitable for degraded lands in arid and semi-arid zones, providing short and long-term economic returns.
The farm-river distance was negatively associated with home garden practices. Hypothetically, proximity to rivers for irrigation in a drought-prone region is important for sustainable farming. In both Bugesera and Rulindo, households access freshwater (rivers) at a distance of 500m. Some respondents in Bugesera reported that they rely on the stagnant water of lakes for household use. In the Shyara area (one of the Bugesera administrative entities), in particular, with a population of 15,856 inhabitants (NISR, 2023), local people hunt for unsafe drinking water in stagnant water left by torrential rains in canyons during the hot and dry seasons. This situation negatively affects human health in a community struggling with sporadic food shortages and household basic needs (decent shelter, food, clean water, education, quality healthcare, and stable employment). Agricultural land’s proximity to water bodies enables flowing water into farms for crop farming (Assefa et al., 2018), such as horticultural production. Akinyemi and Speranza (2022) opined that SSA agricultural land dynamics are driven by various factors, such as farms located far from homes and the proximity to roads and rivers.
- Conclusion
Climate-smart agroforestry is an innovative farming practice, a technology of farming with trees to increase, sustain, and diversify production for food security while adapting to climate change. CSAF is a new farming approach, and research has not yet covered the complexity of CSAF practices, notably the determinants of adoption. The adoption and promotion of CSAF technologies are of vital importance. By employing an advanced methodology, the “Multivariate Probit” (MVP) approach, this study aimed to explore bundles of CSAF practices and their complex interplay with farmer characteristics (demographics, assets, institutions) by going beyond single-practice adoption. The study was driven by two research questions: (i) What are the factors driving the adoption of CSAF practices in the Bugesera and Rulindo agroecological zones of Rwanda? (ii) How do CSAFs correlate and complement each other, guiding more targeted interventions? We analyzed the factors driving the adoption of CSAF practices in the Bugesera and Rulindo agroecological zones of Rwanda using data from a survey of 381 farmers. For this purpose, the factors influencing CSAF adoption were identified, and their marginal effect was determined using the MVP model.
The study found interdependence of CSAF ― a farmer’s adoption of one practice was conditional on the adoption of others. This implies that adopting one CSAF practice can have spillover effects on others, collectively contributing to ecosystem protection, restoration, and sustainability. The research identified significant factors affecting farmers’ adoption choices, such as agroecological zone, altitude, gender, age, farm size, farmer’s experience, radio access, livestock size, and river proximity, showing varying impacts across CSAF practices. Importantly, farmer age, farm size, and experience are predominant factors influencing CSAF adoption. Hence, for the superior adoption of CSAF, focusing on crucial factors associated with CSAF adoption is extremely important. Thus, targeting experienced farmers for CSAF uptake increases the likelihood of adoption and promotion. The positive association of farm size and multipurpose trees and shelterbelts/windbreak practices suggests that land consolidation was likely to favor CSAF adoption and expansion. Conversely, land fragmentation and land size constraints discouraged farmers from adopting CSAF practices for superior outcomes.
To enhance the effectiveness of policies supporting conservation agriculture and rural development, a comprehensive policy framework is needed to promote home gardens, improved fallow, multipurpose trees, tree-crop combinations, and shelterbelts/windbreaks among farmers. Collaborative efforts between government, academia, and industry are essential, emphasizing research, technology transfer, and farmer field schools is crucial to address the complex nature of CSAF adoption. Communication channels, especially radio programs, are vital in sensitizing and disseminating CSAF practices. These findings support the need for tailored interventions that address the unique challenges and opportunities in disparate, contrasting agroecological zones. Collaboration among different sectors can create an enabling environment for CSAF adoption, boosting climate-resilient farming, and sustaining agricultural productivity.
The study advises farmers to jointly adopt various CSAF practices to maximize farm diversity outputs and dispel stereotypes suggesting competition with other crops. The findings encompass physiological, economic, and sociological considerations in CSAF adoption. One limitation of this study is that the data are cross-sectional and might not correctly capture farmers’ characteristics and decisions in response to their adoption of CSAF and variations in a cross-site comparative study. However, recognizing the study’s limitations, further research is encouraged to explore additional factors influencing the adoption and promotion of innovative, climate-smart farming practices, offering insights for more effective strategies.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Ethics statement
Approvals from district mayors were received before fieldwork was undertaken, and informed consent was obtained from all participants in the study. All information was confidential and used for research purposes only. Anonymity was maintained using coded identifiers during analysis to guarantee all personal information would stay private.
Acknowledgments
We are grateful to local leaders in Bugesera and Rulindo, who permitted this study to be carried out in their jurisdictions, district agronomists, extension workers, and farmers, who spared their time to participate.
Data availability statement
Data will be made available on request.
References
Ahmed, M. (2015). Adoption of multiple agricultural technologies in maize production of the Central Rift Valley of Ethiopia. Studies in Agricultural Economics, 117(3), 162–168.
Ahmed, M. (2016). Climate change adaptation strategies of maize producers of the Central Rift Valley of Ethiopia. Journal of Agriculture and Rural Development in the Tropics and Subtropics, 117, 175–186.
Akinyemi, F.O., Speranza, C.I. (2022). Agricultural landscape change impact on the quality of land: An African continent-wide assessment of gained and displaced agricultural lands. International Journal of Applied Earth Observations and Geoinformation, 106, 102644. doi:https://doi.org/10.1016/j.jag.2021.102644
Alavalapati, J.R.R., Nair, P.K.R., Barkin, D. (2001). Socioeconomic and Institutional Perspectives of Agroforestry. In World Forests, Markets and Policies (pp. 71-83). Auburn University.
Ali, A., Erenstein, O. (2017). Assessing farmer use of climate change adaptation practices and impacts on food security and poverty in Pakistan. Climate Risk Management, 16, 183–194. doi:https://www.sciencedirect.com/science/article/pii/S2212096316301164?via%3Dihub
Amadu, F.O., Miller, D.C., McNamara. (2020). Agroforestry as a pathway to agricultural yield impacts in climate-smart agriculture investments: Evidence from southern Malawi. Ecological Economics, 167, 106443. doi:https://doi.org/10.1016/j.ecolecon.2019.106443
Amare, D., Wondie, M., Mekuria, W., Darr, D. (2018). Agroforestry of Smallholder Farmers in Ethiopia: Practices and Benefits. Small-scale Forestry. doi:10.1007/s11842-018-9405-6
ARCOS. (2023). Committed partners launch the pilot phase of one of the biggest community landscape restoration projects in Rwanda: Building sustainable agroforestry value-chains. Kigali, Rwanda: ARCOS Network.
Armengot, L., Barbieri, P., Andres, C., Milz, J., Schneider, M. (2016). Cacao agroforestry systems have a higher return on labor compared to full-sun monocultures. Agronomy for Sustainable Development, 36(4), 36: 70.
Arun, G.C., Jun-Ho, Y. (2020). Perception to Adaptation of Climate Change in Nepal: An Empirical Analysis Using Multivariate Probit Model. Sci Journal, 2(87). doi:10.3390/sci2040087
Assefa, T., Jha, M., Reyes, M., Srinivasan, R., Worqlul, A.W. (2018). Assessment of Suitable Areas for Home Gardens for Irrigation Potential, Water Availability, and Water-Lifting Technologies. Water, 10, 495. doi:10.3390/w10040495
Awazi, N.P., Tchamba, M., Temgoua, L.F. (2021). Climate-Smart Practices of Smallholder Farmers in Cameroon Confronted with Climate Variability and Change: The Example of Agroforestry. Agricultural Research, 83-96. doi:https://doi.org/10.1007/s40003-020-00477-0
Batz, F. J., Peters, K. J., Janssen, W. (1999). The influence of technology characteristics on the rate and speed of adoption. Agricultural Economics, 21(2), 121-130. doi:10.1016/S0169-5150(99)00026-2
Belderbos, R., Carree, M., Diederen, B., Lokshin, B., Veugelers, R. (2004). Heterogeneity in R & D cooperation strategies. International Journal of Industrial Organization, 22(1), 1237–1263.
Bishaw, B., Soolanayakanahally, R., Karki, U., Hagan, E. (2022). Agroforestry Systems, 96(3), 447–451.
Boron, V., Payán, E., MacMillan, D., Tzanopoulos, J. (2016). Achieving sustainable development in rural areas in Colombia: Future scenarios for biodiversity conservation under land use change. Land Use Policy, 59, 27-37. doi:https://doi.org/10.1016/j.landusepol.2016.08.017
Brown, D. G., Robinson, D. T., French, N. H., Reed, B. C. (2013). Land use and the carbon cycle: advances in integrated science, management, and policy. Cambridge: Cambridge University Press.
Brown, S.E., Miller, D.C., Ordonez, P.J., and K. Baylis, K. (2018). Evidence for the impacts of agroforestry on agricultural productivity, ecosystem services, and human well – being in high – income countries: a systematic map protocol. Environmental Evidence, 7(1), 24-46.
Cappellari, L., Jenkins, S. P. (2003). Multivariate probit regression using simulated maximum likelihood. The Stata Journal, 3, 278-294.
Cerdán, C.R., Rebolledo, M .C., Soto, G., Rapidel, B., Sinclair, F.L. (2012). Local knowledge of impacts of tree cover on ecosystem services in smallholder coffee production systems. Agricultural Systems, 110, 119-130.
Chamling, R. S., Gurung, A. (2005). Raising awareness of the impacts of climate change: Initial steps in shaping policy in Nepal. Mountain Research and Development, 35, 316–320.
Chapman, R., Blench, R., Kranjac-Berisavljevic, G., Zakariah, A.B.T. (2003). Rural Radio in Agricultural Extension: the Example of Vernacular Radio Programmes on Soil and Water Conservation in Northern Ghana. Agricultural Research and Extension Network.
Cheo, A.E., Tapiwa, K.A. (2021). SDG 2: Zero hunger: Food security, improved nutrition, and sustainable agriculture. Concise guides to the United Nations Sustainable Development Goals. (W. M. Filho, Ed.) Howard House, Wagon Lane, Bingley BD16 1WA, UK: Emerald Publishing Limited.
Coulibaly, Y.N., Mulia, R., Sanou, J., Zombre, G., Bayala, J., Kalinganire, A., van Noordwijk, M. (2013). Crop production under different rainfall and management conditions in agroforestry parkland systems in Burkina Faso: observations and simulation with WaNuLCAS model. Agroforestry Systems, 88, 13–28.
Danso-Abbeam, G., Baiyegunhi, L.J. (2017). Adoption of agrochemical management practices among smallholder cocoa farmers in Ghana. African Journal of Science, Innovation and Development, 9(6), 717-728.
Deschênes, O., Greenstone, M. (2007). The economic impacts of climate change: evidence from agricultural output and random fluctuations in weather. American Economic Review, 97(1), 354-385.
Donkoh, S., Azumah, S., Baanni, A., Agebase, J. (2019). Adoption of improved agricultural technologies among rice farmers in Ghana: a multivariate probit approach. Ghana Journal of Development Studies, 16(1), 46.
Dorfman, J. (1996). Modeling multiple adoption decisions in a joint framework. American Journal of Agricultural Economics, 78, 547-557.
Dwivedi, R.P., Kareemulla, K., Ramesh, S., Rizvi, R.H., Chauhan, J. (2007). Socio-Economic Analysis of Agroforestry Systems in Western Uttar Pradesh. Indian Research Journal of Extension Education, 7, 18-22.
FAO. (2011). Community listeners’ clubs: Stepping stones for action in rural areas. Brussels, Belgium: FAO-Dimitra.
Garrity, D. (2012). Agroforestry and the future of global land use. In P. G. Nair, Agroforestry – the future of global land use (pp. 21-27). Dordrecht, the Netherlands: Springer Netherlands.
Garrity, D. (2012). Agroforestry and the future of global land use. In P. G. Nair, Agroforestry – the future of global land use (pp. 21-27). Dordrecht, the Netherlands: Springer Netherlands.
Garrity, D., Okono, A., Grayson, M., Parrott, S. (2006). World Agroforestry into the Future. Nairobi, Kenya: World Agroforestry Centre.
Gil, J. D. B., Reidsma, P., Giller, K., Todman, L., Whitemore, A., van Ittersum M. (2019). Sustainable Development Goal 2: Improved targets and indicators for agriculture and food security. Ambio, 48, 685–698.
Glover, E.K., Ahmed, H.B., Glover, M.K. (2013). Analysis of Socio-Economic Conditions Influencing Adoption of Agroforestry Practices. International Journal of Agriculture and Forestry, 3(4), 178-184. doi:10.5923/j.ijaf.20130304.09
Greene, W. (2002). Econometric analysis. New Jersey: 5th Edition.
Greene, W. (2008). The econometric approach to efficiency analysis. In H. L. Fried, The measurement of productive efficiency and productivity growth (p. 656). Oxford, United Kingdom: Oxford University Press.
Harvey, C.A., Gonzales, J.G., Somarriba, E. (2006). Dung beetle and terrestrial mammal diversity in forest, indigenous agroforestry systems and plantain monocultures in Talamanca, Costa Rica. Biodiversity Conservation, 15, 555–585.
Henninger, S. (2013). Does the global warming modify the local Rwandan climate? Natural Science, 5(1), 124-129.
ICRAF. (1993). Strategy to the year 2000. Nairobi, Kenya: ICRAF.
Jagoret, P., Michel-dounias, I. Snoeck, D., Ngnogue, H.T., Male´zieux, E. (2012). Afforestation of savannah with cocoa agroforestry systems : A small-farmer innovation in central Cameroon Afforestation of savannah with cocoa agroforestry systems : a small-farmer innovation in central Cameroon. Agroforestry Systems, 86, 493–504.
Jama, B., Elias, E., Mogots, K. (2006). Role of agroforestry in improving food security and natural resource management in the drylands: a regional overview. Journal of the Drylands, 1, 206-211.
Jara-Rojas, R., Russy, S., Roco, L., Fleming-Muñoz, D., Engler, A. (2020). Factors afecting the adoption of agroforestry practices: insights from silvopastoral systems of Colombia. Forests, 11(6), 648. doi:https://doi.org/10.3390/f11060648
Johnson, R.J., Doye, D., Lalman, D.L., Peel, D.S., Curry Raper, K., Chung, C. (2010). Factors a ecting adoption of recommended management practices in stocker cattle production. Journal of Agricultural and Applied Economics, 42, 15-30. doi:https://doi.org/10.1017/S1074070800003266
Kareemulla, K., Samra, J.S., Rai, P. (2005). Globalization of Industry Oriented Indian Agroforestry: A case study of poplar, eucalyptus, teak and Subabul based agroforestry sectors. South Asia Regional Conference on Globalization of Agriculture in South Asia: Has it made a difference to rural livelihoods, 23-25 March 2005 (p. 36). Hyderabad: CESS, ANGRAU, IFPRI and IAAE.
Kariyasa, K., Dewi, Y.A. (2013). Analysis of Factors Affecting Adoption of Integrated Crop Management Farmer Field School Estimation: (Icm-Ffs) in Swampy Area. International Journal of Food and Agricultural Economics, 1, 29-38.
Kassie, M., Jaleta, M., Shiferaw, B., Mmbando, F., Mekuria, M. (2013). Adoption of interrelated sustainable agricultural practices in smallholder systems: Evidence from rural Tanzania. Technological Forecasting and Social Change, 80(3), 525-540. doi:https://doi.org/10.1016/j.techfore.2012.08.007
Kassie, M., Zikhali, P., Manjur, K., Edwards, S. (2009). Adoption of organic farming technologies: Evidence from semi-arid regions of Ethiopia. Natural Resources Forum, 33, 189– 198.
Khanal, S. (2011). Contribution of Agroforestry in Biodiversity Conservation and Rural Needs Fulflment, Master’s Thesis. Tribhuvan, Nepal: Tribhuvan University-Institute of Forestry.
Khanna, M. (2001). Sequential adoption of site-specific technologies and its implications for nitrogen productivity: A double selectivity model. American Journal of Agricultural Economics, 83(1), 5-51.
Kirby, K. R., Potvin, C. (2007). Variation in carbon storage among tree species: implications for the management of a smallscale carbon sink project. Forest Ecology and Management, 246, 208–221.
Krishnan, V. (2010). Constructing an Area-based Socioeconomic Index: A Principal Components Analysis Approach. Early Childhood Intervention Australia (ECIA). National Convention Centre, Canberra, Australia: ECmap.
Leakey, R.R.B., Weber, J.C., Page, T., Cornelius, J.P., Akinnifesi, F.K., Roshetko, J.M., Tchoundjeu, Z., Jamnadass, R. (2012). Tree domestication in agroforestry: progress in the second decade. In P. G. Nair, The future of agroforestry (pp. 145-173). New York, USA: Springer .
Lin, C.T.J., Jensen, K.L., Yen, S.T. (2005). Awareness of foodborne pathogens among US consumers. Food Quality and Preference, 16(5), 401-412. doi:https://doi.org/10.1016/j.foodqual.2004.07.001
Mai, P. T. (1999). Socio-economic analysis of shifting cultivation versus agroforestry system in the upper stream of lower Mekong watershed in Dak Lak province. MA thesis in Economics of Development. The Hague, The Netherlands: National University-HCMC, Viet Nam and Institute of Social Study (ISS).
Manandhar, S., Vogt, D.S., Perret, S.R., Kazama, F. (2011). Adapting cropping systems to climate change in Nepal: A cross-regional study of farmers’ perception and practices. Regional Environmental Change, 11, 335–348.
Mbata, J. (2001). Determinants of animal traction adoption in traditional agriculture: An application of the multivariate probit procedure to the case of Lesotho. Development Southern Africa, 18(3), 309–325. doi:https://doi.org/10.1080/03768350120069983
Mbow, C., Van Noordwijk, M., Luedeling, E., Neufeldt, H., Minang, A.P., Kowero, G. (2014). Agroforestry solutions to address food security and climate change challenges in Africa. Current Opinion in Environmental Sustainability, 6(1), 61-67.
McNeely, J. A, Schroth, G. (2006). Agroforestry and biodiversity conservation–traditional practices, present dynamics and lessons for the future. Biodiversity Conservation, 15, 549–554.
Mekoya, A., Oosting, S.J., Fernandez-Rivera, S., Van der Zijpp, A.J. (2008). Farmers’ perceptions about exotic multipurpose fodder trees and constraints to their adoption. Agroforestry Systems, 73, 141–153. doi:https://link.springer.com/article/10.1007/s10
Mekuria, M ., Waddington, S. (2004). Institutional and policy support is essential to promote the adoption of soil fertility technologies on maize-based smallholder farms in Southern Africa. The 4th International Crop Science Congress. Brisbane, Australia, 26 Sept. -1 Oct. 2004. Brisbane, Australia.
Mercer, D. E., Frey, G.E. (2014). Economics of Agroforestry. In A. J. Kant. S., Handbook of Forest Economics (pp. 188-209). New York, USA: Earthscan from Routledge.
Mignouna, D.B., Manyong, V, M., Mutabazi, K.D.S., Senkondo, E.M. (2011). Determinants of adopting imazapyr-resistant maize for Striga control in Western Kenya: A double-hurdle approach. Journal of Development and Agricultural Economics, 3(11), 572-580.
Mollier, L., Seyler, F., Chotte, J. L., & Ringler, C. (2017). End hunger, achieve food security and improved nutrition and promote sustainable agriculture: SDG2. In A. S. Stevance, A guide to SDG interactions: from science to implementation (SDG: Sustainable Development Goals) (pp. 31-80). Paris: ICSU.
Montagnini, F. (2006). Environmental services of agroforestry systems. USA: Food Products Press.
Mukundente, L., Ndunda, E., Gathuru, G. (2020). Socio-economic and institutional factors affecting smallholders farmers to adopt agroforestry practices in southern province of Rwanda. International Journal of Agricultural Science and Food Technology, 6(1), 068-074.
Mulwa, C., Marenya, P., Rahut, D.B., Kassie, M. (2017). Response to climate risks among smallholder farmers in Malawi: A multivariate probit assessment of the role of information, household demographics, and farm characteristics. Climate Risk Management, 16, 208–221. doi:https://www.sciencedirect.com/science/article/pii/S2212096317300062?via%3Dihub
Nakabugu, S. B. (2001). The Role of Rural Radio in Agricultural and Rural Development. Translating Agricultural Research Information into Messages for Farm Audiences. Kampala, Uganda: Programme of the Workshop in Uganda.
Nath, T. K., Inoue, M., Myant, H. (2005). Small-scale agroforestry for upland community development: A case study from Chittagong Hill Tracts, Bangladesh. Journal of Forest Research , 10(6), 443-452.
Ndayambaje, J.D., Mohren, G.M.J. (2011). Fuelwood demand and supply in Rwanda and the role of agroforestry. Agroforestry Systems, 83, 303–320.
Ndoli, A., Mukuralinda, A., Schut, A. G. T., Iiyama, M., Ndayambaje, J. D., Mowo, J., G., Giller, K. E., Baudron, F. (2021). On-farm trees are a safety net for the poorest households rather than a major contributor to food security in Rwanda. Food Security, 13(3), 685-699.
Nhemachena, C., Hassan, R., Chakwizira, J. (2014). Analysis of determinants of farm-level adaptation measures to climate change in Southern Africa. Journal of Development and Agricultural Economics, 6, 232–241.
NISR. (2021). Agricultural Household Survey. Kigali Rwanda: Republic of Rwanda.
NISR. (2023). 5th Population and Housing Census. Kigali, Rwanda: National Institute of Statistics of Rwanda (NISR). Retrieved from https://www.statistics.gov.rw/publication/key-figures-5th-rwanda-population-and-housing-census-phc
Nizeyimana, P., Kee-Won, L., Songyong, S. (2018). A study on the classfication of households in Rwanda based on factor scores. ournal of the Korean Data And Information Science Society, 29, 547-555. doi:10.7465/jkdi.2018.29.2.547
Nkamleu, G.B., Manyong, V.M. (2005). Factors afecting the adoption of agroforestry practices by farmers in Cameroon. Small-scale forest economics, management and policy, 4(2), 135–148.
Ntawuruhunga, D., Affognon, H.D., Fiaboe, K.K.M., Abukutsa-Onyango, M.O., Turoop, L., Muriithi, B.W. (2020). Farmers’ knowledge, attitudes and practices (KAP) on production of African indigenous vegetables in Kenya. International Journal of Tropical Insect Science, 40(2), 337-349.
Ntawuruhunga, D., Ngowi, E.E., Mangi, H.O., Salanga, R.J., Shikuku, K.M. (2023). Climate-smart agroforestry systems and practices: A systematic review of what works, what doesn’t work, and why. Forest Policy and Economics, 150, 102937. doi:https://doi.org/10.1016/j.forpol.2023.102937
Ntawuruhunga, D., Twahirwa, M. (2023). Modeling the Influence of Expropriated Farmers’ Determinants on Compensation Payments Using Multiple Regression. Green and Low-Carbon Economy, 00(00), 1-9. doi:https://doi.org/10.47852/bonviewGLCE3202946
Oduniyi, O. S., Tekana, S.S. (2019). Adoption of agroforestry practices and climate change mitigation strategies in North West province of South Africa. International Journal of Climate Change Strategies and Management, 11(5), 716-729.
Ong, C.K., Swallow, B.M. (2003). Water productivity in forestry and agroforestry. Limits and Opportunities for Improvement, 1, 217–228.
Paramita, E., Martini, E., Finlayson, R. (2014). Radio: challenges and potential of using mass media for agroforestry extension in Sulawesi, Indonesia. World Congress on Agroforestry (p. 1). Delhi, India: World Agroforestry Centre.
Patel, N., Sethi, T., Ravisankar, T., Sharma, S., Kumar, R., Pujar, G.S., Arulraj, M., Arunachalam, Handa, A.K., Suresh, R.S., Lesslie, A., Shaik, M., Das, S., Sreenivas, K., Bhanderi, R.J., Prabhu, R., Dhyani, S.K., Singh, R., Roy, P.S., Saigal, S., Kuma. (2024). Greening and Restoration of Wastelands with Agroforestry. New Delhi, India: NITI Aayog.
Piya, L., Maharjan, K.L., Joshi, N.P. (2013). Determinants of adaptation practices to climate change by Chepang households in the rural Mid-Hills of Nepal. Regional Environmental Change, 13, 437–447. doi:https://link.springer.com/article/10.1007/s10113-012-0359-5
Quinion, A., Chirwa, P.W., Akinnifesi, F.K., Ajayi, O.C. (2010). Do agroforestry technologies improve the livelihoods of the resource poor farmers? Evidence from Kasungu and Machinga districts of Malawi. Agroforestry Systems, 80, 457-465.
Rahman, S. A., Rahman, M. F., Sunderland, T. (2014). Increasing Tree Cover in Degrading Landscapes: ‘Integration’ and ‘Intensification’ of Smallholder Forest Culture in the Alutilla Valley, Matiranga, Bangladesh. Small-scale Forestry, 13(2), 237-249.
Rahman, S.A., Rahman, M.F., Codilan, A.L., Farhana, K.M. (2007). Analysis of the economic benefits from systematic improvements to shifting cultivation and its evolution towards stable continuous agroforestry in the upland of eastern Bangladesh. International Forestry Review, 9(1), 536-547.
Raj, A. (2017). Role of Trees and Woody Vegetation in Soil Fertility Enrichment and Food Security in Dryland Agroforestry as a Climate- Smart Agriculture Strategy. International Journal of Tropical Agriculture, 35(4), 1147-1161.
Rutunga, V., Janssen, B.H., Mantel, S., Janssens, M. (2007). Soil use and management strategy for raising food and cash output in Rwanda. Journal of Food, Agriculture and Environment, 5, 434–441.
Saxena, N. (2000). Farm and agroforestry in India – Policy and legal issues. Planning Commission. Government of India.
Schaffnit-Chatterjee, C. (2009). The global food equation: Food security in an environment of increasing scarcity. (S. Schneider, Ed.) D-60262 Frankfurt am Main, Germany: Deutsche Bank Research.
Shackleton, S., Paumgarten, F., Kassa, H., Husselman, M ., Zida, M ., Purnomo, H., Irawati, R.H., Fauzan, A.U., Elati, M. (2012). Forests, gender and value chains. CIFOR Infobrief No. 49. Bogor, Indonesia: CIFOR.
Shahbaz, P.; Haq, S.U.; Boz, I. (2022). Linking climate change adaptation practices with farm technical efficiency and fertilizer use: a study of wheat-maize mix cropping zone of Punjab province, Pakistan. Environmental Science and Pollution Research International, 29(12), 16925-16938. doi:https://doi.org/10.1007/s11356-021-16844-5
Shahbaz, P.; ul Haq, S.; Abbas, A.; Batool, Z.; Alotaibi, B.A.; Nayak, R.K. (2022). Adoption of Climate Smart Agricultural Practices through Women Involvement in Decision Making Process: Exploring the Role of Empowerment and Innovativeness. Agriculture, 12(8), 1161. doi:https://doi.org/10.3390/agriculture12081161
Siebert, A., Dinku, T., Vuguziga, F., Twahira, A., Kagabo, D.M., delCorral, J. (2019). Evaluation of ENACTS-Rwanda: A new multi-decade, highresolution rainfall and temperature data set—Climatology. nternational Journal of Climatology, 1-17. doi:. https://doi.org/10.1002/joc.6010
Sunderland, T., Abdoulaye, R., Ahammad, R., Asaha, S., Baudron, F., Deakin, E., Duriaux, J.Y., Eddy, I., Foli, S., Gumbo, D., Khatun, K., Kondwani, M., Kshatriya, M., Leonald, L., Rowland, D., Stacey, N., Tomscha, S., Yang, K., Gergel, S., Van Vianen, J. (2017). A methodological approach for assessing cross-site landscape change: Understanding socio-ecological systems. Forest Policy and Economics, 84, 83-91.
Tega, M., Bojago, E. (2023). Farmer’s Perceptions of Agroforestry Practices, Contributions to Rural Household Farm Income, and Their Determinants in Sodo Zuria District, Southern Ethiopia. International Journal of Forestry Research, 19. doi:https://doi.org/10.1155/2023/5439171
Tengnas, B. (1994). Agroforestry extension manual for Kenya. Nairobi, Kenya: International Centre for Research in Agroforestry.
Tesfaye, W., Seifu, L. (2016). Climate change perception and choice of adaptation strategies: Empirical evidence from smallholder farmers in east Ethiopia. International Journal of Climate Change Strategies and Management, 8, 253–270. doi:https://www.emerald.com/insight/content/doi/10.1108/IJCCSM-01-2014-0017/full/html
UNDP. (2016). UNDP Support to the Implementation of Sustainable Development Goal 13. New York, NY, USA: United Nations Development Program (UNDP).
United Nations. (2015). Global Sustainable Development Report, 2015 edition. New York, USA: United Nations.
United Nations. (2016). The Sustainable Development Goals Report 2016. New York, USA: United Nations.
United Nations. (2020). Goal 2: Zero Hunger. Sustainable development goals. doi:Retrieved from https://www.un.org/ sustainabledevelopment/hunger/
Venkatesh, A., Mohapatra, K.P., Arunachalam, A., Rajkhowa, D.J., Bharali, S., Ngachan, S.V. (2014). Agroforestry and Food Security. In A. L. Raj, Agroforestry: Theory and Practices (pp. 34-53). India: Scientific Publishers.
Wen, Y., Zhou, X., Yu, S., Zhu, H. (2018). The predicament and countermeasures of development of global Eucalyptus plantations. Guangxi Sciences, 25(2), 107-116.
Wonnacott T. H., Wonnacott R. J. (1967). Introductory Statistics. John Wiley & Sons Inc.
Wooldridge, J. (2003). Diagnostic testing. In B. Baltagi., A Companion to Theoretical Econometrics (p. 730). Cowley Road, Oxford, OX4 1JF, UK: Blackwell Publishing Ltd.
Wu, J.J., Babcock, B.A. (1998). The choice of tillage, rotation, and soil testing practices: Economic and environmental implications. American Journal of Agricultural Economics, 80(3), 494–511.
Xu J, Mercado A, He J., Dawson, I. (2013). An Agroforestry guide for field practitioners. Kunming, China: The World Agroforestry Centre, East Asia.