John Libbey Eurotext

Science et changements planétaires / Sécheresse


Modeling and participatory farmer-led approaches to food security in a changing world: A case study from Malawi Volume 24, issue 4, Octobre-Novembre-Décembre 2013


See all figures


Auteur(s) : Sieglinde Snapp1, Rachel Bezner Kerr2, Alex Smith1, Mary Ollenburger3, Wezi Mhango4, Lizzie Shumba5, Tinkani Gondwe5, George Kanyama-Phiri4

1 Michigan State University Department of Plant, Soil and Microbial Sciences Kellogg Biological Station East Lansing, MI USA

2 Cornell University Department of Development Sociology Ithaca, NY USA

3 Wageningen University Plant Production Systems Group P.O. Box 430 6700 AK Wageningen The Netherlands

4 Lilongwe University of Agriculture and Natural Resources Bunda College Lilongwe Malawi

5 Ekwendeni Hospital Soils Food & Healthy Communities Project Ekwendeni Malawi

Reprints: S. Snapp

In a rapidly changing world, new and growing sources of uncertainty are posed by climate change. More than ever, there is a need for capacity building to support adaptation on the part of farmers. Smallholder farmers living in the semi-arid tropics are particularly at risk to climatic variability such as poorly timed dry spells, extended drought and excess rainfall that can cause flooding events (Funk et al., 2008). Poor soils add to vulnerability, through insufficient capacity to buffer water availability, and a constant nutrient deficient environment. Degraded soils often interact with, and exacerbate conditions when rainfall is variable. Education is key to helping farmers cope with these challenges. Ecological intensification options and technologies can support enhanced food security in this variable environment, but only if scientists and agricultural development workers are engaged in full partnerships with farmers and rural communities. Here we discuss how participatory approaches and simulation modeling can contribute towards capacity building for sustainable agriculture.

Smallholder farmers face labour, land and resource constraints. What supports or drives adoption, and adaptation, of technologies has rarely been fully understood. Engaging with farmers through participatory research and education is one way to both understand and support farmers in developing innovations suited to these constraints. Stability of production, meeting a diversity of requirements, labour and land saving traits, and locally-preferred taste and storage traits are often valued above high yield of specific crops (Mugwe et al., 2009; Snapp et al., 2010). ‘Better bet’ or ‘plausible bet’ technology options can be developed through participatory action research to support farmer innovation (Chambers, 1994; Snapp et al., 2002). Participatory research, linked to modeling and analysis of farming system livelihoods, are approaches that support building farmer capacity, and co-learning in an iterative manner among all participants (Bezner Kerr et al., 2007).

It is important for scientists and farmers to work together to assess performance and risk associated with technological options potentially used to support sustainable agriculture (Snapp et al., 2002; Schulz et al., 2003; Snapp et al., 2003a). Food security is often a key attribute that must be met when introducing new technical options, without enhancing the risk of crop failure, excess labour requirements or uneconomic returns. It can be challenging to introduce alternative plant species to meet food security requirements, as producing sufficient amounts of a staple cereal, such as maize, is a top priority for many small-scale farmers,. Growing mixtures of diverse legume species with maize has been one of the major technology options put forward as a means to address multiple objectives (Gilbert, 2004; Snapp et al., 2010). In Northern Malawi both male and female farmers show strong preferences for technologies that produce food for both sale and nutrition (Mhango et al., 2013). Similarly in West Africa there appears to be a strong preference among smallholder farmers for food legumes, relative to soil improving and multipurpose legumes (Morse and McNamara, 2003; Schulz et al., 2003).

Performance of crops and farming systems interact strongly with weather, and the risk of climatic variability can be assessed through crop simulation modeling, although the success of model prediction will depend on many factors. Over a decade of experience has helped fine-tune a number of crop simulation models for assessing on-farm performance in sub-Saharan Africa (Shamudzarira and Robertson, 2002; Robertson et al., 2005; Chikowo et al., 2008). Farmer participatory research can benefit from starting with the assessment of options that might fit local community requirements for a range of species, novel varieties or animals, and management options that are environmentally sound and support production of diverse dietary and marketable products - often called ‘best bet’ options. For example, in Kenya, opportunity for adoption of legume-based technologies has been assessed in a systematic manner, leading to a range of ‘best bet’ options for farmers to try out (Ojiem et al., 2007).

There are many examples of capacity building, adaptation, and adoption of sustainable agriculture innovations by smallholders living in the semi-arid tropics (SAT), and lessons can be learned. We draw upon an example from Northern Malawi, at a site where we have been engaged in participatory action research with local communities since 2000 (Bezner Kerr et al., 2007; Bezner Kerr et al., 2011). We have found that farm families in this area, similar to many farming communities across the SAT, place a high priority on food security and child nutrition, with gender influencing assessment of technologies, as well as other factors. Diverse goals may occur within a farm family (e.g., male vs. female farmers) as well as across communities and regions, so different options may be relevant for different groups (Tittonell et al., 2010).

Here we present insights learned from working for over a decade in the Ekwendeni region of Northern Malawi, including a novel assessment of a farmer-led approach linked to simulation modeling of cropping system performance for a range of households. The Ekwendeni region of Northern Malawi has a highly variable unimodal rainfall pattern (figure 1), and is representative of many smallholder farm sites in Southern Africa. We explore how farmer preferences, adaptations and innovations interact with model predictions to influence food security associated with crop diversification options in a highly variable climate.

This project has been based on a close partnership with ‘Soil Food and Healthy Communities’ (SFHC) project of Ekwendeni Hospital and scientists from several universities and disciplines – including faculty from University of Malawi agricultural sciences, Michigan State University soil and crop ecology, University of Western Ontario geography and a nutritionist from HealthBridge Canada. We have worked closely for over a decade with farmer research groups, Ekwendeni SFHC, Malawi extension and international crop scientists, among others, to support farmers experimenting with diversification (Bezner Kerr et al., 2007). Malnutrition and food insecurity is an acute problem in Northern Malawi. There are many socioeconomic and policy issues involved, but the fundamental lack of diverse foodstuffs is part of the problem in rural communities. Malawian agricultural systems are dominated by maize with as much as 60 to 70% of crop land devoted to producing maize on smallholder farms across the country (Snapp et al., 2002). While maize is a productive cereal, and a highly preferred food type by most Malawians, growing a wider range of crops provides advantages in terms of spreading risk, improving diets, and addressing market opportunities (Snapp et al., 2010).

The crop diversification options being tested, and fine-tuned, by farmer research groups in Ekwendeni vary from species that are primarily useful for enhancing soil fertility such as fish-poison-bean (Tephrosia vogelii) and velvet bean (Mucuna pruirens), to those that are primarily useful for food production such as pigeon pea (Cajanus cajan) and soybean (Glycine max). One of the specific crops that has emerged as a ‘better bet’ option when grown in rotation with maize at this site is velvet bean, which is a green manure crop that produces copious amounts of leafy residues to build soil fertility, but also produces a grain that is used occasionally for food after time-intensive boiling and preparation. This legume was popular with about 15% of farmers engaged with trying crop diversification, primarily those households with more resources and access to labour such as village leaders (Bezner Kerr et al., 2007; Snapp et al., 2010). Achieving much higher popularity has been food legumes, grown alone or in mixtures, particularly pigeon pea which produces food, but has multiple uses in addition (Mhango et al., 2013). Pigeon pea is a shrubby crop that can be grown for one or two years as an intercrop, and as a rotational crop, and is well suited to intercrop with maize as it initially grows very slowly and thus the plant imposes minimal competition for light, moisture and other resources with the main crop. Shorter-statured food legumes have also proved popular, including groundnut (Arachis hypogaea), cowpea (Vigna unguiculata) and soybean, grown on their own or in mixtures with pigeon pea and maize. In the last few years there has also been growing interest in diversifying grain crops and adding roots and tubers such as cassava (Manihot esculenta) and sweet potatoes (Ipomoea batatas), to increase the range of crops to draw on in the face of rainfall unpredictability.


There are two sources of data for this paper. The first includes data from a combination of a cross-sectional survey (n=435), interviews and informal observations of ongoing participatory research carried out by hundreds of farmers over the past ten years. The Ekwendeni Hospital that researchers at this case study site collaborate with initially selected villages to participate in the project based on expressed interest, high levels of food insecurity and location within the hospital catchment area (Bezner Kerr et al., 2007; Bezner Kerr et al., 2011). Farmers could then volunteer to be in the project and test out different legume intercrops as a means to improve soil fertility, food security and nutrition. By 2005, there were 3800 farmers who had received a small amount of legume seeds, some training from fellow farmers, and participated in various educational activities, including field days, recipe days or farmer apprenticeship programs. Interviews and observations were conducted annually with participating farmers. The research team designed a survey in August 2005 to assess dietary practices, nutritional knowledge, and farmer experiences with legumes. Ten enumerators were trained in the survey, pilot tested and revised the questions and interviewed 435 respondents, resulting in 419 valid surveys: 213 participating farmers, and 206 control households (102 who lived in participating villages and 104 farmers who lived in non-participating villages). The survey was approved by Cornell University's Institutional Review Board for Human Subjects, and included a full informed consent, attention to confidentiality and privacy.

The second source of data for this paper is from crop simulation modeling outputs (Ollenburger, 2012), which we used to derive estimates of whole farm food security. We focused on the pigeon pea mixed cropping systems with maize for this exercise. Pigeon pea is an important legume as it provides multiple benefits through a unique growth habit, as it grows slowly initially which makes it an ideal intercrop, then it grows into a bush during the dry season after maize is harvested. These traits allow it to build soils through a long period of nutrient accumulation, biological nitrogen fixation and producing large amounts of soil-conserving leaf litter. In addition, pigeon pea provides nutritious food products in the form of green pods for vegetable use, and a dry grain that is high in protein (Snapp et al., 2003b). Maize was chosen along with pigeon pea to explore mixed systems through simulation modeling, as maize is the staple cereal crop of the region. Model outputs were used to evaluate the potential of cropping systems to meet household food needs, based data from a survey of farming households in Ekwendeni. Climate records from a site in Ekwendeni were used to run long-term simulations and assess food security risk over time for a range of households.

Crop simulation

Modeling was conducted using the Agricultural Production Systems Simulator (APSIM), parameterized with data from on-farm trials conducted in Ekwendeni, Northern Malawi (Chikowo et al., 2008; Ollenburger, 2012). We simulated crop growth, yield and soil processes for a sole maize crop, a maize/pigeon pea rotation, and a maize + pigeon pea intercrop, employing meteorological data collected at Zombwe, Malawi (in the Ekwendeni region) between 1945 and 2011 (figure 1). Soil properties were based on data collected from participating farmers’ fields (Ollenburger, 2012). Soils used in this study were sandy loam texture with 79% sand and 14% clay. Topsoil organic carbon content was 0.58%. A small dose of nitrogen fertilizer input was used in the simulation, with 24 kg N/ha applied whenever maize was planted in any of the three systems. The sole and intercrop systems required fertilization every year based on continuous maize presence, whereas the rotation required half as much fertilization as no maize was present half the time. In rotation systems, yields were annualized by assuming that farmers planted two staggered rotations, each occupying half their land. Simulations were run with crop coefficients for the hybrid maize variety MH-17, which was already parameterized for APSIM and is popular Malawi maize variety, and a generic long-duration pigeon pea variety. These simulations were compared against data from experimental plots in Ekwendeni, Malawi (Mhango, 2011; Ollenburger, 2012).

For each calendar year between 1956 and 2010, APSIM simulations were run for each system in its first through tenth consecutive year of establishment. This was accomplished by running ten multi-year simulations with their start dates staggered by one year, and with all simulations re-set every 10 years (Ollenburger, 2012). We initially compared the probability of meeting food needs between early and late years of establishment, but as few differences were apparent, we pooled all years of establishment for the final analysis.

APSIM does not simulate crop losses from pests, including both pathogens and animal pests. While the model is capable of simulating losses due to weed competition, insufficient data were available for parameterization of weed model components for this case. When simulated maize yields were compared with actual maize yields from comparable experimental plots in Ekwendeni, Malawi, which were exposed to pests, modeled yields were typically within one standard deviation above measured yields (Ollenburger, 2012). The comparison of modeled to measured yields does not account for post-harvest crop loss, which is a major concern for farmers in Ekwendeni. We do not have adequate data to estimate crop losses due to factors extrinsic to the model, but based on values from the literature we have reduced all modeled yields by 33% (Schulten, 1975; Mhango et al., 2013; Ollenburger, 2012).


We selected 12 case study households from the full set of interviewed households on which to model crop production and calculate food needs. These case study households were purposely selected to represent a broad range of household types, based on survey data collected from 200 farming households in Ekwendeni, Malawi, during May of 20101. These data included demographic information, indicators of wealth, amount of arable land farmed by each household, and information about farming practices and crop varieties used.

Case study households were chosen to represent the full range of wealth scores among surveyed households: including the least wealthy, the wealthiest, and 10 households that fell between these, evenly distributed across household wealth scores. We then calculated the probability that these case study households would meet all food needs, as described below. The wealth of households was scored based on a point system developed through participatory wealth ranking exercises, drawing as well from other studies in Malawi that included socioeconomic variables in their analysis (Smale and Phiri, 1998; Hotz and Gibson, 2001). This system awarded points for livestock owned, the quality of construction materials used in the family house, commercial items owned, and access to an irrigated vegetable garden. We also report additional survey information for each case study household, including head-of-household gender, agricultural practices, and sale of crops.

Meeting food needs

In order to calculate the probability of each household meeting its food needs under climatic conditions for each year between 1956 and 2010, we first calculated the calorie and protein needs for each household based on the number of children under 13, adults 13-69, and seniors 70 and over. We then obtained human calorie requirements from the FAO Human Energy Requirements report (Food and Agriculture Organization, 2004), assuming 1.9x BMI for adults (moderate to high activity level) and moderate activity level for children. Protein requirements were obtained from the WHO Protein and Amino Acids in Human Nutrition report (World Health Organization, 2007). Finally, we determined the calories and protein produced per hectare for each run of the crop simulation model, combining the maize and pigeon pea grain yield produced. Calorie content and protein content of the grain maize and pigeon pea were obtained from a Malawi food composition database2.

The probability of each household meeting its calorie needs for each cropping system was calculated as:


Where Phc is the probability of household h meeting its calorie needs under cropping system c, yci is the total calorie yield per hectare from cropping system c in simulation i, ah is the area of arable land farmed by household h, in hectares, and rh represents the total calorie requirements for household h. A similar calculation was used to determine a household's probability of meeting their protein needs. Simulations of continuous maize, maize/pigeon pea rotation, and maize + pigeon pea mixed cropping in each system's first through tenth year of establishment during every calendar year from 1956 through 2010 were used in these calculations.

All calculations and data management functions were performed in R or in standard spreadsheet software (R Development Core Team, 2011).


Farmer adoption

Drawing from the 2005 survey, legumes were a common crop grown during the rainy season by many farmers: 75% grew groundnuts, 60% grew soybeans, 53% grew common beans, 43% grew pigeon peas, 21% grew cowpea, 20% grew velvet bean and 11% grew fish-poison-bean. The area devoted to legume production was relatively small: an average of 4,046 m2 for soya beans, 2,056 m2 for groundnuts, and 1,123 m2 for pigeon pea. Farmers grew legumes for multiple reasons, including food, income, improving soil fertility and improving children's nutrition (figures 2, 3 and 4). Groundnut, for example, was typically grown as a source of income, since it is an important cash crop, but it was also grown as a food that can be added to the staple maize dish (a porridge) to improve family nutrition. Pigeon pea, in contrast, was typically grown to improve soil fertility, with additional interest in the crop as a food source. Finally farmers ranked ‘to add to children's porridge’ or ‘to improve children's health’ as the most important reason that they grew soybean, while food was the second reason (figure 4), with income and soil fertility as the least important reasons.

In contrast to the typical farming practice of sole cropping in this area, many farmers increasingly intercropped their legumes. The most common intercrop combination was maize and soybean, followed by pigeon pea and groundnuts, pigeon pea and maize and finally pigeon pea and soybean. The majority of farmers reported burying crop residue after harvest as a means to improve soil fertility and observed dark green (72%) or green (24%) maize leaves in fields where legumes had been grown the previous year, and understood that using legume residues was a type of income generation strategy through reducing fertilizer costs. In addition, farmers sold some of their legume grain, particularly groundnuts and soybeans, and used the income for various household expenses.

Farmers reported that project staff and other participating farmers were the main source of information for learning about using legumes to improve soil fertility. In addition, the majority of participants reported sharing this information with an average of 11 other farmers. During interviews, farmers also talked about additional knowledge gained through this project - beyond soil fertility benefits of growing multiple legumes on one field. Farmers described the following: saving labour during planting and weeding; getting food at different times of the year; having diverse dietary options; and gaining different sources of income. Farmers additionally reported experimenting with various combinations of legumes (e.g., cowpea and pigeon pea) and other crops (e.g., cassava and soybean) and tried different spacing and growing times. For example, some farmers chose to increase cowpea production as this legume provides a useful leafy vegetable during the period of food shortages and also provides a grain legume later in the year. Farmers also reported intercropping cowpea with maize, sorghum or pigeon pea successfully.

We found that farmers developed a number of innovative management techniques, and one that proved most popular was intensification of legume presence using a three year system. Initially, a doubled up system of pigeon pea and groundnuts, or soybean, was grown in year 1, then farmers cut back the branches of the pigeon pea after pods were harvested, and planted maize amongst pigeon pea for an intercrop system in year 2. Farmers then harvested both crops, incorporated remaining pigeon pea residues completely, and grew maize in year 3. By allowing pigeon pea to grow for multiple years, farmers were able to get reasonably sized shrubby trees that could be used for fuelwood, substantially greater soil fertility benefits, and two years of nutritious pigeon pea grain. This adaptation highlights farmers’ interest in multiple ecosystem services, and long-term sustainability of production systems (Snapp et al., 2003a; Snapp et al., 2010).

Farmers observed that soils with legume residues incorporated over several seasons were able to retain more moisture during periods of low precipitation when compared to fields without multiple years of legume crop and organic material addition. As farmers have became more concerned about ways to reduce risk from what they perceived to be increasingly unreliable rainfall patterns, growing intensified legume mixtures with maize was reported as an important type of climate change adaptation strategy. Farmers have also cited increased climatic risk as their reason to test drought tolerant local maize varieties, increase cassava production and generally diversify their cropping system.

Modeling food security

The proportion of years that case study households were able to meet calorie and protein requirements was generally high, particularly for the two legume-diversified cropping systems that were modeled (figures 5 and 6). The intensity of legume presence each year varied in the cropping systems from zero (sole maize), to 50% (maize/pigeon pea rotation), to 100% (maize + pigeon pea intercrop) of the time. The pigeon pea + maize intercrop was best able to meet household food needs, meeting calorie needs in 73 to 100% of the years simulated for 10 out of 12 case study households, and meeting protein needs in 70% to 100% of the years simulated for all twelve households (figures 5 and 6). The rotation system was almost as successful, performing as well as the intercrop in the majority of cases, but failing for households with large family size relative to their land base (figures 5 and 6, table 1). Sole maize was consistently the poorest performer, and consistently met calorie requirements for only half the households (figure 5).

Table 1 Characteristics of the 12 selected case study households.

Household Children under 13 Adults Seniors 70 and over Acres arable land Wealth score Household head gender Sell maize Use fertilizer Intercrop maize and legumes
1 1 6 0 0.76 2 F N N N
2 0 1 0 1.62 5 F N N N
3 3 4 0 1.92 6 M N Y Y
4 5 2 1 1.11 8 M N Y N
5 2 6 1 1.32 8 M N Y N
6 6 7 0 1.52 10 M N N N
7 4 2 0 2.43 11 M Y Y Y
8 4 2 0 1.21 12 M N Y N
9 3 5 0 0.61 12 M N Y Y
10 2 5 0 0.91 14 M N Y Y
11 4 2 0 1.21 16 M N Y N
12 1 4 1 2.43 23 M Y Y Y

Food requirements were determined in this study for each household based on the number of adults and children, while the food each household produced in a given year under a given cropping system was a function of farm size. Therefore those households with a large family size, and the accompanying requirements for calories and protein, were at risk of not achieving sufficient food production to meet requirements in years with unfavourable growing conditions. This ‘at risk’ situation was exacerbated for households with a small land base. Households 1, 9 and 10 had large family size in relation to their land base (table 1), and these households experienced protein and calorie deficits in most years with the notable exception of simulations using the intercrop system (figures 5 and 6).

Wealth was not closely associated with either family size or land size, thus in our study it did not influence the proportion of years in which households met food security needs. However, in this study we did not take into account the ability of better-resourced households to purchase food or hire more labour, mitigating food insecurity. The model outputs are thus a starting place for further explorations of food security issues related to climate change, farmer preferences and innovations.


This project took a farmer-led approach which fostered farmer experimentation and information sharing. The promotion of legumes, intercropping and residue management has been promoted by many organizations in Malawi and other parts of Africa, in order to improve soil fertility and quality, as well as provide nutritious food and extra income (Snapp et al., 1998; Giller et al., 2011). However, success has been mixed, with limited adaptation and adoption of legume-based innovations. In our case study, farmers were given a small amount of seeds and some training from other farmers, then left to experiment with seed spacing, plant combinations, and agronomic practices. Farmers were an integral part of the research team by being actively involved in planning with researchers and a farmer research group, monitoring progress, training other farmers, and leading seed management that included setting up community seed banks. The farmer-to-farmer approach was clearly evident from the survey results on sources of information. There were multiple lines of evidence indicating that co-learning and capacity building occurred, both among households that interacted with the project intensely and among those that were not formally part of the project. For instance, we found that new genotypes of seeds sourced initially from the project were widely shared, with many ‘control’ households planting pigeon pea in doubled-up legume cropping systems arrangements and using improved methods for residue management. The majority of respondents reported burying pigeon pea residue to improve soil fertility which was a dramatic shift from the historic practice of burning or clearing residue prior to planting, and indicated that they had learned this information from fellow farmers in the SFHC project (Bezner Kerr et al., 2007). Interestingly, burning and clearing continue to be the most widely used residue management approach among households in non-participating villages.

The project promoted a wide range of legume species, including agroforestry species as well as food legumes. Farmer reported preferences, and adoption patterns that we observed in Ekwendeni, all highlighted that legumes were primarily prized for their food production properties. This was the number one use for groundnut and pigeon pea (figures 2 and 3), whereas infant feeding was the number one use for soybean, with food production being ranked a close second (figure 4). Farmers were also exposed to legumes that provided substantial soil fertility benefits, such as the agroforestry maize-relay system using fish-poison-bean and a green manure maize rotation using velvet bean, yet only a small minority of farmers showed long-term interest in these legumes (Bezner Kerr et al., 2007). One encouraging outcome of this project was that a novel legume for this area, pigeon pea, was adopted by thousands of Ekwendeni farmers as this legume provided the desired combination of food plus soil fertility benefits (Mhango et al., 2013).

We focused crop simulation modeling on the production potential of pigeon pea and maize crops in order to assess food security implications for a range of households from the project area. Under different climatic growing conditions, based on decades of rainfall and temperature records from Ekwendeni, we found that pigeonea + maize intercrop systems were highly likely to produce sufficient calories and protein for smallholder farm households, with 83% of case study households being able to meet calorie needs at least 73% of the time. This stands in stark contrast to monoculture maize, which only supported households meeting calorie needs 50% of the time. The rotation system was intermediate in performance, according to these model outputs. We note, however, that the rotation was associated with modest fertilizer inputs, as no N fertilizer inputs were applied to pigeon pea in the simulation, which has important sustainability implications relative to the other systems (Snapp et al., 2010). Overall, model simulations were run with the assumption that modest doses of nitrogen fertilizer would be available to apply to maize, 24 kg N/ha, based on historic fertilizer use across much of Malawi.


In contrast to an emphasis in the agricultural research community on monoculture cereal production to meet caloric needs (Smale and Heisey, 1997; Denning et al., 2009), our model output and assessment by farmers suggested that crop diversification may be central to efforts to enhance food security. In particular, maize-pigeon pea mixtures, and other legume intercrops were valued by farmers in Ekwendeni. Modeling cropping system productivity in relation to food security requirements showed that intercrops of maize and pigeon pea were more reliable for meeting caloric and protein needs than sole cropped maize, even for households with large families. By expanding the focus of agricultural research beyond simple grain yield, we can address food security and other multipurpose needs that diversified farming systems address, including ecosystem services and livelihoods.


We have been fortunate to receive support through long-term funding from the International Development Research Centre, the McKnight Foundation, United States Agency for International Development (USAID) ‘Feed the Future’, the Canadian Food Grains Bank and Presbyterian World Service and Development. We would like to sincerely thank Danielle Zoellner-Kelly for editorial input and suggestions that improved this manuscript.