EFFECTS OF MAIZE COWPEA INTERCROPPING ON YIELD STABILITY AND PRODUCTION RISK IN CENTRAL MALAWI: A MODELING STUDY By Alex A. Smith A THESIS Submitted to Michigan State University i n partial fulfillment of the re quirements for the degree of Crop and Soil Sciences Master of Science 2014 ABSTRACT EFFECTS OF MAIZE COWPEA INTERCROPPING ON YIELD STABILITY AND PRODUCTION RISK IN CENTRAL MALAWI: A MODELING STUDY By Alex A. Smith Smallholder farmers in Malawi face growing challenges from increasingly variable weather and scarce resources. We used the Agricultural Production Systems Simulator (APSIM) to investigate yield stability and production risk in maize cowpea intercrop and maize and cowpea monocrop systems at three locations in central Malawi. This was done as part of an ongoing farmer participatory research project, with a household survey and field trials providing context for the modeling study. Models were calibrated and validated based on field trials, and cropping systems were simulated f or 26 seasons. Simulations were run in two soils at each site, under unfertilized conditions and with the addition of 69 kg N / ha. We determined calorie and protein yields of intercrop and monocrop systems in relation to household needs, and used stabil ity analysis to contrast yields across a range of environmental conditions. Calorie and protein yields tended to be higher in intercrops than in monocrops, except at one location where unfertilized maize yields were suppressed in intercrop with cowpea. C alorie and protein yields of intercrops also tended to be more variable than monocrops, largely due to the high performance of intercrops under favorable conditions. Maize cowpea intercrop systems demonstrated a greater chance of meeting household calorie and protein needs than maize and cowpea monocrops in 83% of cases. This suggests that maize cowpea intercropping is generally advantageous for smallholder farmers in Malawi, but yields may be suppressed in environments where cowpea is a strong competitor with maize. This study is unique in that it addresses risk to farmers and compares system performance across a range of growing conditions and seasons. iii ACKNOWLEDGEMENTS Thanks to Dr. Sieglinde Snapp for taking me on as a student and providing unwavering support through many trials and transitions. Thanks also to Dr. Laura Schmitt - Olabisi, Dr. Karen Renner, Dr. Regis Chikowo, Dr. Wezi Mhango and Dr. Rachel Bezner Kerr for their guidance and support. Special thanks go to Mary Ollenberger for patiently int roducing me to crop modeling. Thanks to Edward Mzumara, Michele Hockett, Isaac Jambo, Dziwani Kambauwa, Elian Mjamanda, Emmanuel Jambo, Kondwani Khonje and Miriam Mhango for being wonderful friends and co - workers. And thanks as well to Emmanuel Mbewe, Be nart Msukwa, Mark Freeman, and Wilson Ndovie for sharing their expertise. Particular thanks and acknowledgements go to Amos Ganizani, Jimmy Dinesi, Jakena Phiri and Blessings Kadzembuka, whose efforts in the field made everything possible. My deepest tha nks to all the farmers who took part in the Africa RISING project, for their time and energy, for their welcoming kindness, and for their belief that a better future is possible. Finally, thanks to the United States Agency for International Development and to Michigan State University for their financial support. iv TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES LIST OF ABBREVIATIONS and SYMBOLS INTRODUCTION Climate variability, food security and agricultural intensification in Malawi The ecology of intercropping and evaluation of intercrop systems Maize cowpea intercrop systems Crop simulation modeling and the Agricultural Production Systems Simulator QUESTIONS AND HYPOTHESES Questions Hypotheses MATERIALS AND METHODS Research conte xt Survey data collection Agronomic trials Soil sampling and analysis Crop modeling Model calibration and validation Exploratory modeling Statistical analysis A nalysis of a gronomic trials Stability analysis Risk analysis RESU LTS Farming household characteristics and practices Agronomic trials Model calibration and validation Modeled yields Stability analysis Risk analysis DISCUSSION CONCLUSIONS APPEND ICES Appendix A: Description of soils at the Africa RISI NG Malawi 2013 intervention sites v i vi i vii i 1 1 5 9 12 16 16 17 18 18 19 20 21 24 24 27 30 30 30 30 33 33 35 35 38 4 1 4 5 4 8 5 7 5 8 59 v Appendix B: All parameters used in exploratory simulation models Appendix C: Outputs from models in which soil parameters were not re - set annually WORKS CITED 63 68 7 1 vi LIST OF TABLES Table 1: Study locations classified by water availability . Table 2: Soil properties measured in experimental trials and used in subsequent modeling Table 3: Parameters manipulat ed during model calibration . Table 4: Soil C and N used in high organic matter and low organic matter soils in long - term simulation models. Table 5: Household composition and landholding based on surveys of farming households at the study locations. Tabl e 6: Comparison of mean yields in agronomic trials at all study locations in the 2012 - 2013 growing season. Table 7: R 2 values for linear regressions used in stability analysis plots. Table A. 1: Description of soils from Africa RISING mother trial s. Table A. sites. Table B. 1: Soil parameters used in exploratory crop simulation models. Table B. 2: Management parameters used in exploratory cro p simulation models. Table B. 3: Other parameters used in exploratory crop simulation models. 19 23 26 29 31 35 45 59 61 63 65 67 vii LIST OF FIGURES Figure 1: Comparison of cumulative precipitation at the study locations with precipitation a t nearby meteorological stations. Figure 2: Percent of households growing legume crops, and percent growing legumes in intercrop with maize . Figure 3: Percent of households growing sole maize, sole legumes, and maize legume intercrops, and percent applyi ng mineral N fertilizer and manure to these systems. Figure 4: Calibration and crossvalidation of crop simulation models. Figure 5: Modeled grain yields of cropping systems from 1980 2005. Figure 6: Modeled biomass yields of cropping systems from 1980 2005. Figure 7 : Stability analysis of calorie and protein yields from modeled cropping systems. Figure 8 : Probability that average households will meet calorie and protein needs with their own production. Figure C. 1: Modeled grain yields of c ropping systems from 1980 2005 in which soil parameters were not re - set annually, allowing for cumulative effects. Figure C. 2: Modeled biomass yields of cropping systems from 1980 2005 in which soil parameters were not re - set annually. Figure C. 3: Probability that average households will meet calorie and protein needs with their own production; contrast of simulations in which soil parameters were reset annually and simulations in which parameters were not reset annually. 28 34 3 4 36 39 40 4 4 4 7 68 69 70 viii LIST OF ABBREVIATIONS and SYMBOLS APSIM: Agricultural Production Systems Simulator C: Carbon N: Nitrogen P: Phosphorus P (as used in probability notation): Probability 1 INTRODUCTION Cli mate variability, food security and agricultural intensification in Malawi Smallholder farmers practicing rain fed agriculture in sub - Saharan Africa must cope with present - day insecurity and a future rife with uncertainty. Many rural households are cu rrently food - insecure, and food demand is expected to increase in coming decades due to a growing population and changing patterns of food consumption (Thornton et al. , 2011) . At the same time, climate change will alter patterns of temperature and rainfal l in sub - Saharan Africa, and may cause many areas to develop climate regimes with no present - day analogue. Without adaptation in maize ( Zea mays ) based agriculture, yields of this principal food crop are expected to grow less stable in coming decades (Lob ell et al. , 2008; Thornton et al. , 2011) . With the threat of climate change and increasing demand s placed on agroecosystems, farmers will need to adapt to new conditions and stresses. Farmers practicing rain fed agriculture have already developed strate gies for coping with present - day climate stresses, and these strategies can serve as a starting point for adapting to an uncertain future (Thomalla et al. , 2006) . However, smallholder farmers tend to have very limited access to resources, and this may for ce them to prioritize short - term food production over investments in longer - term agroe cological stability (Snapp et al. , 2002b) . This highlights the importance of working with farmers to develop adaptation strategies that fit within the context of farmers constraints (McDougall and Braun 2003, Snapp et al. 2003). By supporting adaptation strategies intended to improve crop production and food security under present - day environmental variability and stress es continue adapting in an uncertain future (Thomas and Twyman, 2005; Twomlow et al. , 2008) . Due to chronic food insecurity in sub - Saharan Africa and the threats of climate change and population growth, there have been frequent calls for a new African gree n revolution 2 (Toenniessen et al. , 2008; Snapp et al. , 2010) . The first green revolution, which combined fertilizer - responsive hybrid grains with the intensive use of fertilizers, herbicides and pesticides, met with little success in Africa. Per - capita fo od production in Africa has actually declined in recent decades, while increasing in most other regions of the world (Toenniessen et al. , 2008) . This is due at least in part to the high cost of agricultural inputs in Africa, in conjunction with incomes of less than $1 / day for many smallholder farmers (Kerr and Kolavalli, 1999) . If future efforts at intensification are to meet with greater success, they must be commiserate with the serious constraints faced by many smallholder farmers in Africa. Proposal s for the form that this new green revolution should take are varied. Some proponents focus on the increased use of fertilizers and other inputs, made possible by heavy government subsidies (Denning et al. , 2009) . However, exclusive reliance on subsidize d inputs can prove problematic when those subsidies dry up due to changes in national policy (Chirwa et al. , 2006) . It has therefore been argued that, in order to provide sustainable benefits to smallholder farmers, the African green revolution must be fo unded on ecological intensification and agricultural biodiversity (Snapp et al. , 2010) . Efforts in this direction have focused mainly on the incorporation of legumes and organic amendment into cereal cropping systems, along with the modest use of fertiliz ers and other inputs (Vanlauwe et al. , 2010) . Most legumes provide a supplemental source of fertility through biological nitrogen fixation, and can benefit subsequent grain crops (Chikowo et al. , 2004) . When used in conjunction with modest fertilizer app lications, legumes can increase the benefit from each kilogram of N applied (Snapp et al. , 2010) . Intensification efforts based on the integration of legumes into cereal cropping systems may therefore prove more accessible to resource - constrained smallhol der farmers. 3 Smallholder farmers in Malawi must cope with severe constraints as they adapt their agricultural systems to present and future conditions. Slightly over half of all Malawians earn less than a dollar a day, and a similar proportion of Malawian children are stunted due to malnutrition (Wood and Mayer, 2006) . In drought years farmers can face partial or total crop failure, rendering them unable to meet household food needs through their own production (Kamanga et al. , 2010) . Soil fertility is t ypically low, limiting the productivity of low - input agriculture (Snapp, 1998) . The effects of these biophysical constraints are compounded by shrinking landholdings, high food prices, eroding social support networks, and the deterioration of natural area s that rural people rely upon to buffer food supplies in times of scarcity (Chinsinga, 2004; Scholes and Biggs, 2004; OECD/FAO, 2012) . Nevertheless, many Malawian farmers continuously experiment with new cropping patterns and other agricultural strategies to make the most of the resources available to them (Peters 2002). Farmers in Malawi have traditionally planted cereal and legume crops together in intercrop systems, and this practice remains prevalent in smallholder fields (Shaxson and Tauer, 1992; Pete rs, 2002) . Intercropping systems, like other forms of crop diversification, spread risk among multiple crops, so that the failure of a single crop does not mean a failure to produce any food or income (Trenbath, 1999) . Smallholder farmers frequently cite intercropping and other forms of mixed cro pping as an important strategy for coping with climate variability (Hassan and Nhemachena, 2008; Ozor and Cynthia, 2010) . Intercropping systems can be particularly attractive to farmers in Malawi because they ens ure that a large portion of their limited arable land can be devoted to production of the staple crop, maize (Shaxson and Tauer, 1992; Snapp et al. , 2002b) . In Malawian smallholder agriculture, crops are planted into ridges constructed with hand hoes. In traditional cropping systems planting ridges tend to be spaced relatively widely, 4 but in recent years more intensive spacing has been heavily promoted and widely adopted by farmers (Denning et al. , 2009) . This represents a widespread transition from a re latively conservative traditional approach , to an approach intended to maximize maize production. This transition may be paralleled by a decline in intercropping, as a recent nationwide survey (Jones et al. , 2014) reports decreased rates of intercropping compared with past surveys (Ngwira et al. , 1990) . Despite the continuing prevalence of intercropping in Malawi and across sub - Saharan Africa, many major research efforts have focused almost exclusively on crop rotation as a means of integrating legumes int o cereal - based systems (Giller et al. , 2009) . Similarly, agricultural extension agencies in sub - Saharan Africa have tended to discourage intercropping (Byerlee and Heisey, 1996) , though extension in Malawi has included intercropping treatments in some lar ge - scale trials (Snapp et al. , 2010) . Critical evaluation of the risks and benefits of intercropping in African smallholder systems can shed light on the place of this practice in efforts towards agricultural intensification. One of the chief aims of leg ume based intensification is to improve soil fertility over time through the production of N - rich biomass (Snapp et al. , 1998; Thierfelder et al. , 2012) . However, farmers experiencing chronic food insecurity may be forced to prioritize immediate survival over long term considerations (Bryceson and Fonseca, 2006) . Farmers in Malawi have expressed concern about any loss of maize production due to legumes, and while soil fertility is cited as a concern, it comes after food production and crop sales (Snapp et al. , 2002b) . Food insecurity can force farmers to use cropping practices that are known to degrade the soil when they believe that those practices are the best option for meeting their immediate food needs (Murage et al. , 2000) . Therefore, when evaluati ng options for sustainable intensification in 5 smallholder systems, it is important to prioritize short - term food production alongside long - term improvements in soil fertility. The ecology of intercropping and evaluation of intercrop systems A large body o f ecological theory and research indicates that biodiversity tends to be linked to improved ecological productivity and stability (Hooper et al. , 2005) . While much of this research has been conducted in natural systems, the same principals apply to croppi ng systems. Two general mechanisms can allow intercrops to be more productive than monocrops. The competitive production principal states that different species will exploit slightly different resource pools, making inter - specific competition less intens e than intra - specific competition (Vandermeer, 1995) . This is observed in cases where intercrops achieve greater water use efficiency by drawing water at different times in the year (Sileshi et al. , 2011) , or from different depths in the soil profile (Ogi ndo and Walker, 2005) . The competitive production principal also comes into play when one intercrop partner is capable of fixing N from the atmosphere rather than relying entirely on N from the soil (Fukai and Trenbath, 1993; Chikowo et al. , 2004) . Furth ermore, interplanted crops can actively facilitate the growth of their partner crops. This can occur through the provision of biologically fixed N in the form of N - rich crop residues (Snapp et al. , 1998) , through the creation of microclimates that limit wa ter loss (Morris and Garrity, 1993) , and through the suppression of weeds and insect pests (Skovgard and Pats, 1996) . While crops grown together in intercrop can be subject to a number of productivity - increasing interactions, they are also subject to int er - specific competition. Intercrop partners compete with one another for limited resources including water, light, and nutrients. Competition for water can limit the productivity of intercrops in dry years, resulting in higher 6 yield variability in respon se to climatic conditions (Fukai and Trenbath, 1993; Ollenberger, 2012) . While biological N fixation limits the degree of competition for N in cereal legume intercrops, N fixing legumes can still be limited by N competition in intercrop systems (Ghosh et al. , 2006) . Competition for light can also determine whether crop yields are suppressed in intercrop, particularly in cases where a short statured crop is grown with a tall - statured crop such as maize (Keating and Carberry, 1993) . Due to the interplay o f productivity - increasing and productivity - limiting interactions in intercrop systems, intercrops will only demonstrate a yield advantage over monocrops in situations where inter - specific competition is less intense than intra - specific competition, or wher e competition between crops is out - weighed by facilitative interactions (Vandermeer, 1995) . Intercropping can take numerous forms combining two or more annual crops, perennial crops, or some combination of the two. For intercrops involving two annual cr ops with similar growth periods, intercrop systems can be distinguished into two basic forms (Fukai and Trenbath, 1993) . Intercrops can be substitutive, with crop plants of one species replaced with plants of another. Substitutive intercrops have a total plant density similar to the plant density of either crop in monoculture . Alternately , intercrops can be additive, with crop plants of both species planted at the same density that would be used in monocrop. Additive intercrops have a higher total plant density than either crop i n mono culture . These differences in plant density can alter the intensity of competition in intercrop systems, affecting the dynamics of the system (Rusinamhodzi et al. , 2012) and its relative advantage over monoculture (Watiki e t al. , 1993) . Additive intercropping is commonly practiced by smallholder farmers in Malawi, and is particularly prevalent among households with limited land per - capita (Shaxson and Tauer, 1992) . Farmers in neighboring Mozambique reported a preference f or additive intercrop systems over 7 substitutive systems due to their higher yield potential (Rusinamhodzi et al. , 2012) . This preference for additive systems is supported by a modeling study of maize cowpea intercrops, which found that additive or semi - ad ditive systems produce optimal economic yields (Adiku et al. , 1998) . Intercropping cereals with legumes has been found to increase the productivity of cereal cropping systems across many locations and climatic conditions in sub - Saharan Africa (Waddington e t al. , 2007; Snapp et al. , 2010; Thierfelder et al. , 2012) . This is largely driven by legume s capacity to meet a portion of their own N requirements through biological N fixa tion, and their production of N - rich biomass which can contribute to the N nutri tion of subsequent cereal crops (Chikowo et al. , 2004; Vesterager et al. , 2008) . However, cereal legume intercropping does not necessarily offer an advantage under all cond itions. The benefits of legume intercrops may diminish in input - intensive cropping systems, where the N needs of cereal crops are supplied by mineral fertilizer (Searle et al. , 1981; Carberry et al. , 1996; Sileshi et al. , 2011) . In contrast, legumes may also contribute little under excessively harsh growing conditions. Legumes grown i n extremely poor soils may grow poorly and contribute little N - rich biomass to the soil (Chikowo et al. , 2007; Mhango et al. , 2012) . Finally, some arrangements of crops in intercrop may produce a yield advantage over monocrops, while others do not (Watiki et al. , 1993) . We must therefore carefully evaluate the appropriateness of intercrop systems on a situation by situation basis and improve our understanding of how the environment interacts with intercrop system performance. When contrasting monocrop and intercrop systems, it is necessary to find a basis for comparison between a system with one crop and a system with two or more. The yield of a single crop cannot fully express the productivity of an intercrop system . Numerous methods are 8 used for compa ring the productivity of monocrops and intercrops. Perhaps the most commonly used index for comparison is the land equivalent ratio (LER), which is the ratio of the land necessary to produce a given yield of two crops in monocrop to the land necessary to produce those same yields in intercrop. Unfortunately , the LER may be biased in favor of intercrops in situations where the two intercrop partners differ in productivity, as in most maize legume systems (Connolly, 1986) . Monocrop and intercrop productivi ty are also frequently compared in terms of economic yield (Lightfoot et al. , 1987; Adiku et al. , 1998) , though the results are dependent on fluctuating crop prices. Total calorie and protein yiel d of the cropping system has been proposed as an alternativ e basi s for comparing intercrop s and monocrop s (Trenbath, 1999) . Calorie and protein yields are likely to be particularly relevant to smallholder farmers who produce some or most of their own food . Yield stability from season to season is a vital factor t o consider when evaluating smallholder cropping systems. Because smallholder farming households often produce barely enough to meet their subsistence needs , yield stability is likely to impact food security in a variable and changing climate (Giller et al . , 2006; Kamanga et al. , 2010) . Whether intercropping systems are more or less stable than monocrops is a matter of debate, and may vary based on the crops involved (Fukai and Trenbath, 1993; Altieri, 2004) . Relative yield stability may also depend on the arrangement of crops in the cropping system (Shumba et al. , 1990; Rusinamhodzi et al. , 2012) . Stability can be conceived of in several ways. Frequently, stability is thought of as the degree of variation around the mean, often expressed as the coeffici ent of variation (Waddington et al. , 2007) . Stability can also be thought of as responsiveness to varying environmental conditions. This is seen in stability analysis or adaptability analysis, in which yields are plotted 9 against an environmental index ra nging from the most favorable to the least favorable conditions (Lightfoot et al. , 1987; Hildebrand and Russell, 1996) . Yield stability is important, but it is not co nsistently low yields may be considered stable, but may be entirely undesirable for farmers. yield and variability (Hildebrand and Russell, 1996) . Measures of p yields, and their need to produce yields that are stable over time. Risk analysis measures the chance that a cropping system will meet or exceed a minimum required outcome (Hildeb rand and Russell, 1996) . By measuring the frequency with which the minimum outcome has been met in years past, risk analysis gives the likelihood that the outcome will be met in the future. This minimum required outcome can be a minimum economic yield (L ightfoot et al. , 1987; Yamoah et al. , 2000) , or the minimum food production required to meet household needs (Kamanga et al. , 2010) . By measuring the chance that a farming household will produce enough to meet their needs, risk analysis provides a criteri on that is directly relevant to food security and livelihood sustainability. Maize cowpea intercrop systems Cowpea ( Vigna unguiculata ) is an annual legume common to smallholder cropping systems in sub - Saharan Africa. It was brought into cultivation in Af rica, and wild relatives can be found across the continent (Padulosi and Ng, 1997) . Cowpea cultivars display a variety of growth habits, ranging from fully determinate to highly indeterminate. It is valued for its edible foliage as well as for its grain, and can serve as a food source in the hungry period before harvest 10 time (Nielsen et al. , 1997) . Cowpea is relatively drought tolerant, and is capable of extracting water from the soil at soil water concentrations below the lower limit for maize (Adiku et al. , 1998) . Unfortunately, cowpea is also susceptible to many insect pests and diseases, and yields may be depressed unless pesticides or biological control methods are used (Jackai and Adalla, 1997) . In African smallholder cropping systems, cowpea is t ypically grown in intercrop with maize, sorghum, pearl millet, and other cereal crops, though monocropped cowpea is common in some regions (Mortimore et al. , 1997) . Maize and cowpea exhibit complementary use of water and N resources in intercrop systems. Cowpea in intercrop with maize has been found to supply up to 72% of its N needs through biological N fixation, though this percentage is depressed under P limited conditions (Vesterager et al. , 2008) . Maize cowpea intercrops have also been found to exh ibit higher water use efficiency than either crop in monocrop (Hulugalle and Lal, 1986) . In intercropping systems maize has been found to withdraw the majority of water under wet conditions, while cowpea is capable of withdrawing soil water under conditio ns too dry for maize, resulting in complementary water use (Adiku et al. , 2001) . The productivity of maize in intercrop with cowpea relative to maize in monoculture may depend on the fer tility inputs applied to the system. Additive m aize cowpea intercrops have been found to produce maize yields that are equivalent to maize monocrops, while also producing cowpea, when no N is added to the system (Ofori and Stern, 1987; Rusinamhodzi et al. , 2012) . However, under fertilized conditions, maize yields in maize cowpea intercrops can be sharply depressed when compared with maize monocrops (Ofori and Stern, 1987; Watiki et al. , 1993) but see (Rusinamhodzi et al. , 2012) . Similar trends have been found in other maize legume 11 intercrops, including maize pigeonpea and maze leucaena (Kamanga et al. , 2010; Sileshi et al. , 2011) . Fertility inputs can also be a factor determining intercrop advantage when the productivity of the entire system is considered . Unfertilized maize cowpea intercrops have been found to produce a n LER above 1.5 (an LER of one indicates that an intercrop and its respective monocrops are equally productive) (Ofori and Stern, 1987) , and produce higher economic yields than their respective monocrops (Carberry et al. , 1996) . These studies have found t hat when N is applied to the system, mai ze cowpea intercrops achieve an LER closer to one (Ofori and Stern, 1987) and lose their economic advantage over monocrops (Carberry et al. , 1996) . The superior performance of maize legume intercrops under unfertili zed conditions is likely due to biological N fixation by legumes, allowing them to remain productive in conditions where maize is N limited (Vesterager et al. , 2008) . Intercrop advantage can also result from the deposition of N rich residues in the soil, and the availability of that N to subsequent maize crops (Chikowo et al. , 2004) . This advantage could disappear under conditions where N is not limiting. However, it should be noted that intercrop advantage is not always restricted to unfertilized system s, and in some cases maize cowpea intercrops have been found to achieve maize yields similar to monocrop regardless of the N rate applied, while producing cowpea in addition (Rusinamhodzi et al. , 2012) . The productivity and yield stability of maize in inte rcrop with cowpea can also depend on the way that plants are arranged in the system. A long term study of substitutive maize cowpea intercrop systems in Zimbabwe found depressed maize yields in intercrop with cowpea, but improved yield stability (Waddingt on et al. , 2007) . Likewise, in a comparison of additive and substitutive systems in Mozambique, maize in substitutive intercrop with cowpea yielded less 12 than maize in monocrop in both favorable and unfavorable seasons (Rusinamhodzi et al. , 2012) . However , maize in additive intercrop with cowpea produced yields equivalent to monocrop in a favorable season, but lower yields than monocrop in an unfavorable season. These studies suggest that substitutive intercrop systems may provide improved yield stability and experience decreased maize yields relative to monocrop, while additive systems can achieve comparable maize yields but may suffer reduced yield stability. Where the productivity of the entire cropping system has been considered, the stability of ma ize cowpea intercrops relative to monocrops has varied from one study to another. In a long term modeling study of maize cowpea cropping systems, the economic yield of intercrops was found to be more stable over time than the yield of either monocrop (Adi ku et al. , 1998) . In contrast, both additive and substitutive maize cowpea intercrop systems in Zimbabwe demonstrated LERs below 1 in an unfavorable season, and LERs above 1 in a more favorable season, indicating reduced yield stability compared to maize monocrop (Shumba et al. , 1990) . Furthermore, in sorghum cowpea systems in Botswana, intercrops produced higher economic yields than monocrops, while yield stability was similar in monocrop and intercrop systems (Lightfoot et al. , 1987) . This emphasizes that stability alone is not a sufficient criterion for evaluating intercrop systems, and a system may be of greater benefit to farmers without being more stable. Crop simulation modeling and the Agricultural Production Systems Simulator Process - based cr op simulation models offer the ability to explore the performance of cropping systems under a wide range of environmental conditions. These models are based on equations describing physical and biological processes such as the mineralization and 13 immobiliz ation of N and the movement of water through the soil profile (Probert et al. , 1998) , and physiological processes involved in crop growth (Carberry et al. , 1989) . Using climate and soil parameters measured in the field, along with crop parameters determin ed through experimental trials, they simulate the interaction of crops, soils, and the atmosphere from one day to the next (Keating et al. , 2003) . The Agricultural Production Systems Simulator (APSIM) was developed to simulate semi - arid rain fed cropping systems in Australia, and has since been used extensively in sub - Saharan Africa (Keating et al. , 2003) . Based on the CERES maize model, it has been expanded to simulate a wide range of crops and cropping systems (Carberry et al. , 1989) . The capacity of APSIM to simulate maize and cowpea has been extensively tested against data from tropical agroecosystems , including systems in sub - Saharan Africa (Carberry et al. , 1989; Adiku et al. , 1995; Keating et al. , 2003) . APSIM is relatively unique among process based crop simulation models in its capacity to simulate intercrop systems (Malezieux et al. , 2009) . Though the model has no spatial component, it employs an arbiter routine to partition resources including light, water, and N between intercrop partners ( Carberry et al. , 1996) . F eatures relevant to the simulation of intercrop systems include simulation of water and nutrient uptake at different layers of the soil profile, simulation of light availability and the effects of shading, simulated plant residue deposition and subsequent N mineralization, and simulation of the effects of soil surface litter on soil water (Keating et al. , 2003) . The ability of APSIM to simulate maize cowpea intercrops has been tested using field data from sub - Saharan Africa (Adiku et al. , 1995) . While models such as APSIM are powerful, they also have serious limitations. APSIM does not include many important ecological factors such as the action of pests and disease, 14 spatial interactions between components of the agroecosystem, and the dynamic, responsive management of cropping systems by farmers (Keating et al. , 2003) . Despite these limitations, APSIM and related models remain powerful tools which have been extensively used to eval uate risk in cropping systems (Nelson et al. , 2002) , explore the possi ble effects of climate change (Asseng et al. , 2013) , and evaluate agricultural technologies in collaboration with farmers (Whitbread et al. , 2010) . APSIM has been used to simulate smallholder cereal legume cropping systems in Malaw i (Robertson et al. , 2005; Ollenberger, 2012) , and to simulate the chance that farming households will meet their food needs under a variety of cropping systems (Snapp et al. , 2014) . In order to ensure that models provide a credible representations of actu al systems, it is necessary to compare the behavior of simulated cropping systems to the behavior of actual systems (Xiong et al. , 2008) . This is typically achieved through the process of calibration and validation. In model calibration, coefficients are adjusted to cause simulated yields to conform to observed yields. Calibration must be undertaken with caution if too many model parameters are adjusted, or if parameters are adjusted outside of reasonable bounds, this process calls into doubt the model (Sinclair and Seligman, 2000) . Model validation is the use of a model to simulate an environment that it was not calibrated for, tion is strongest when performed across a range of environmental conditions that represent the full scope of interest for the investigation (Sinclair and Seligman, 2000) . Calibration and validation alone are not sufficient to ensure that crop models will play a constructive role in the development of intensified cropping systems that can benefit smallholder farmers (Cash et al. , 2003) . To make the greatest contribution to the process of sustainable 15 development, models must be salient (relevant to actual systems and situations) and legitimate (acceptable to stakeholders in those systems), as well as being credible (providing a good approximation of the behavior of actual systems). While credibility can be established by contrasting modeled and actual syst ems (Xiong et al. , 2008; Asseng et al. , 2013) , salience and legitimacy depend on interactions between modelers and stakeholders in the systems being modeled (Cash et al. , 2003) . To be of use to farmers, modeling must be conducted in a context where farmer s and other stakeholders are involved in the research process, and therefore have a basis for trusting its results (van den Belt, 2004) . 16 QUESTIONS AND HYPOTHESES Questions We used simulation modeling to explore yield stability and production risk in mai ze cowpea intercrop systems and maize and cowpea monocrops at three study locations in central Malawi. This study was conducted as part of a farmer participatory research project established in central Malawi in 2012, building on over a decade of research on legume - diversified maize cropping for sustainable intensification (Kamanga et al. , 2010; Snapp et al. , 2010) . We addressed the following questions: - Do maize cowpea additive intercrops show increased calorie and protein yields in their first year of es tablishment when compared with maize and cowpea monocrops? Is this influenced by N fertilization? - Do maize cowpea intercrops show reduced yield stability across seasons and soils compared with maize and cowpea monocrops? - C an maize cowpea intercrops incre ase a typical farming calorie and protein needs through their own production , compared with maize and cowpea monocrops? 17 Hypotheses - Maize cowpea additive intercrops will show higher mean calorie and protein yields than maize monocrops in the absence of N inputs, but not under N fertilized conditions. - Maize and cowpea calorie and protein yields will be less stable across seasons and soils in maize cowpea intercrop systems compared with maize and cowpea monocrops. - We do not ha ve enough information to predict whether intercrops will increase a typical g calorie and protein needs through their own production. 18 MATERIALS AND METHODS Research context This study was conducted as part of the Africa RISING Malawi research initiative, exploring strategies for sustainable intensification of maize - based cropping systems in central Malawi based on long - term farmer participatory research in the country (Snapp et al. , 1998) . Our modeling investigation involved t hree locations in central Malawi in the 2012 2013 growing season. These were selected to represent three agroecological zones: high elevation with high water availability, mid elevation with medium water availability, and low elevation with moderate to low water availability (Table 1). Africa RISING utilized a participatory research framework known as the mother / baby trial design (Snapp et al. , 2002a) . Mother / baby trials situate agronomic research on working smallholder farms and involve farmers in the research process in order to make the work as relevant as possible to farmers. At each study location, two clusters of villages were identified, and participation was solicited from farmers in those villages. Participating farmers worked together, a contrasting a range of cropping systems. In addition, farmers tested two or more cropping their own needs and growing conditions. In total there were two agronomic trials (mother trials) per study location, accompanied by many baby trials. In this study we focus on the researcher - designed, farmer and researcher managed mother trials, employi ng them as a basis for crop simulation modeling. Soils at the Kandeu study location are a mix of orthic ferralsols and chromic luvisols. Soils at Lintipe are ferric livisols, and soils at Golomoti are a mix of eutric fluvisols and eutric cambisols (Lowo le, 1984) . The surface soil at trial one at Kandeu was sandy loam, while the soil 19 at trial two was sandy clay loam. At Lintipe, the surface soil at trial one was clay, and at trial two, clay loam. The surface soil at Golomoti trial one was loam (Table 2 ). The percent total carbon in surface soils at our study locations ranged from 2.3% at Lintipe trial one, to 0.6% at Kandeu trial two. Soil pH at our study locations was generally low, and surface soil pH ranged from 5.13 at Lintipe trial two to 4.5 a t Golomoti trial one. Phosphor us content of soils varied drastically across our study locations, with Bray extractable P in surface soils ranging from 72.5 ppm at Kandeu trial one, to 4 ppm at Kandeu trial two. Table 1: Study locations classified by water a vailability. Classification is based on an index of mean annual rainfall over 30 years at each si te and mean annual temperature (Chikowo 2013, personal communication). Site name: District Altitude Water availability Latitude Longitude Kandeu Ntcheu 909 m Medium - 14.6292 34.5975 Linthipe Dedza 1236 m High - 14.2058 34.1009 Golomoti Dedza 554 m Low - 14.4386 34.6039 Survey data collection We collected data on household composition and farming practices in each study location through a survey conducted in the sum mer of 2013. At each location , the survey included at least 40 households participating in the Africa RISING project, at least 20 local control households who were not participating in Africa RISING but were co - located in the same villages with farmers who were, and at least 20 distant control ho useholds in villages 10 km or more from the intervention villages. The survey included 81 households at the Kandeu study location, 83 households at the Lintipe study location, and 80 households at the Go lomoti study location. Farmers were asked to provide detailed information on cropping practices in each field they managed, including all inputs applied, all crops grown and their planting arrangement (intercrop or sole crop), and an assessment of soil fe rtility in each of their fields. 20 A gronomic trials Five experimental trials grown in the 2012 2013 season were used to calibrate and validate models in this study. There were two trials at the Kandeu study location, two at Lintipe, and one at Golomot i. Each trial included up to 12 treatments, replicated three times each in 5 m x 5 m plots and arranged in a randomized complete block design. Treatments were selected for each location to reflect a mix of current farmer practice, along with sustainable intensification strategies judged to be appropriate for the local environment. At trial one in Kandeu, treatments included unfertilized maize monocrop, maize monocrop receiving the current recommended fertilizer rate for maize in Malawi (a basal applicatio n of 23 kg N and 21 kg P / ha plus a topdress of 46 kg N / ha), maize cowpea intercrop receiving the recommended fertilizer rate, and cowpea monocrop receiving a reduced fertilizer rate (a basal application of 12 kg N and 11 kg P / ha). The maize variety planted in this trial was Sc403, and the cowpea variety planted was Sudan 1. Trial two at Kandeu included both maize treatments and the maize cowpea intercrop as above, but no sole cowpea. The maize variety planted at Kandeu two was DKC 8053, and the cow pea variety planted was IT82E - 16. At Lintipe, trials one and two included unfertilized maize, maize at the recommended fertilizer rate, and sole cowpea at the reduced fertilizer rate. The maize variety planted at Lintipe trial one was PAN 53, and the cow pea variety was Sudan 1. At Lintipe trial two, the maize variety planted was DK 9089 and the cowpea variety was IT82E - 16. At Golomoti trial one, treatments were the same as the Lintipe trials. Sc403 maize was planted at Golomoti, along with Sudan 1 cowp ea. All varieties were sourced from development organizations active in Malawi. Insect pests were controlled in cowpea in all trials using dimethoate , an insecticide sometimes used by farmers . 21 All crops were planted in December 2012. Crops were plant ed on ridges formed by hand - hoe, following local practice. Maize densities ranged from 40,000 plants / ha to 67,000 plants / ha depending on the trial. Cowpea densities ranged from 67,000 plants / ha to 167,000 plants / ha, depending on the trial. Diff erences in plant population density between trials were due to local ridge spacing (between 0.6 and 1 meter apart), which differed in response to environmental yield potential, and variation in the spacing of cowpea on the ridge (between 10 and 15 cm apart ). Maize was spaced at 25 cm along the ridge. While maize is less closely actively promoted by extension and has been widely adopted by farmers in Malawi (Denn ing et al. , 2009) . Intercrops were fully additive, such that in each trial where a maize cowpea intercrop was planted, maize density in intercrop was equal to maize density in monocrop, and cowpea density in intercrop equaled cowpea density in monocrop. Soil sampling and analysis We collected soil profile data from the first agronomic trial at each study location. Soil pits were dug to a depth of 120 cm approximately one month prior to the time of sampling. At the time of sampling, a new face was crea ted using a spade. We then took samples from the fresh face at depths of 0 - 15 cm, 15 30 cm, 30 60 cm, 60 90 cm, and 90 120 cm. At each depth, we took one 98 cm 3 soil core, hammered into the side of the pit and carefully removed to keep the core intact. In addition, we took one sample of disturbed soil at each depth. At Kandeu, the pit could only be dug to a depth of 90 cm due to impenetrable clay subsoil, and 90 - 120 cm samples were not taken. 22 Additional samples were taken from three random p oints in all agronomic trials. Samples were taken with a hand hoe at three depths: 0 15, 15 30 and 30 60 cm. Samples from each depth were composited, resulting in one 0 15 cm, one 15 30 cm, and one 30 60 cm sample from each trial. At Lintipe trial two, hand hoe samples were taken at depths of 0 15 and 15 30 cm only. We also sampled soils from fields belonging to farmers participating in the Africa RISING project and interviewed in the baseline 2013 household survey. Surveyed farmers rank each site, we purposively selected among fields located near the agronomic trials three fields that fields per location. Samples were taken from three random points within each field at depths of 0 15 cm and 15 30 cm using a hand hoe. The three samples taken from each depth were composited, resulting in one sample from 0 15 cm and one from 15 30 cm from each field. At each field, we inquired which crops had been grown in the 2012 / 2013 season. We took GPS coordinates at all sampled fields using a Garmin eTrex 10 GPS unit. All disturbed soil samples were passed through a 2mm sieve at the Bunda College of Agriculture soil lab, and rocks and large pieces of organic matter were discarded. Samples were air - dried, and a subsample was shipped to Michigan State University for physical and chemical analysis . Undisturbed soil cores were oven dried for 48 hours at 130 o C at the Bunda College of Agriculture, and dry soil from the cores was weighed to determine bulk density. 23 Soil texture was determined from sieved soil using the hydrometer method (Kellogg Biol ogical Station LTER, 2008) . Soil pH in distilled water at a 1:2 soil to water ratio was determined (Hendershot et al. , 1993) , using a Metler Toledo SevenEasy S20 pH meter. A subsample of each soil was pulverized using a shatter mill for chemical analysi s. Total carbon and total nitrogen were determined from pulverized soil using a using a Carlo Erba NA1500 SeriesII Combustion Analyzer (Kellogg Biological Station LTER, 2003) . Available P was determined from pulverized soil by the Michigan State Universi ty Soil and Plant Nutrition Laboratory using the Bray extraction method (Bray and Kurtz, 1945) . Table 2: Soil properties measured in experimental trials and used in subsequent modeling. Bulk density is given as BD; % C and % N indicate percent total C and N in the soil; Bray P indicates Bray - extractable phosphorus . Depth (cm) % Sand % Silt % Clay BD (g/cm 3 ) % C % N pH Bray P (ppm) Kandeu trial 1 0 - 15 75 17 9 1.5 1.0 0.09 5.0 72.5 15 - 30 74 13 13 1.4 0.7 0.06 5.4 42.5 30 - 60 68 17 16 1.3 0.8 0.06 5.2 27.5 60 - 90 40 20 40 1.4 1.5 0.12 4.9 9.5 Kandeu trial 2 0 - 15 59 12 29 NA 0.7 0.06 5.0 4.0 15 - 30 50 10 40 NA 0.4 0.05 4.8 2.0 30 - 60 47 11 43 NA 0.2 0.04 4.9 2.0 Lintipe trial 1 0 - 15 22 29 48 1.4 2.3 0.13 4.9 13.5 15 - 30 19 26 55 1.5 2.2 0.12 4.9 12.5 30 - 60 18 24 58 1.3 2.0 0.11 5.0 8.5 60 - 90 6 16 78 1.4 0.7 0.05 5.1 2.5 90 - 120 5 19 76 1.3 0.4 0.04 5.2 2.5 Lintipe trial 2 0 - 15 52 17 31 NA 1.5 0.10 5.1 9.5 15 - 30 38 16 46 NA 1.7 0.10 5.2 3.0 Golomoti trial 1 0 - 15 52 22 26 1.4 0.9 0.06 4.5 56.5 1 5 - 30 43 20 37 1.4 0.5 0.05 4.2 20 30 - 60 34 19 47 1.3 0.1 0.04 4.9 12 60 - 90 27 22 51 1.5 0.2 0.05 4.3 16.5 90 - 120 29 21 50 1.7 0.2 0.02 4.2 12 24 Crop modeling Model calibration and validation Crop models in APSIM version 7.5 were parameterized using soil data collected from the agronomic trials at each of the study locations. For the first trial at each location, data on soil C, soil N, pH, bulk density and texture from depths of 0 - 15, 15 - 30, 30 60, 60 90, and 90 - 120 cm were used in model param eterization (Table 3). For Kandeu trial one, sampled soils from 60 90 cm were used to parameterize the 60 90 and 90 120 cm soil layers. For the second trial at each location, data on soil C, soil N, pH and texture from depths of 0 - 15, 15 - 30 an d 30 - 60 cm were used to parameterize models. At Lintipe trial two, data from 0 - 15 and 15 - 30 cm only were collected. Data from trial 1 were used to parameterize the 60 90 and 90 120 soil layers at trial two. In addition, bulk density data from t rial one were used to parameterize trial two Soil water at saturation, field capacity, and the 15 - bar lower limit were determined using pedo - transfer functions based on % sand, % clay, % soil organic matter, and bulk density (Saxton and Rawls, 2006) (Tabl e 3). We determined soil organic matter as a function of total soil carbon (Sleutel et al. , 2007) . Additional soil parameters, including soil water at air dry, crop LL, crop PAWC, crop XF, initial soil water, and initial mineral N were obtained from a li brary of generic soil profiles for APSIM (APSIM Initiative, 2013) . For each trial, we selected a generic soil profile that was as similar as possible to the actual soil. For the first Kandeu trial, sampled soils from 60 90 cm were used to parameterize the 60 90 and 90 120 cm soil layers (Mabapa et al. , 2010) . 25 For the second trial at each location, models were parameterized based on data from soils collected from depths of 0 - 15, 15 - 30 and 30 - 60 cm. At Lintipe data from 0 - 15 and 15 - 30 cm only were collected. Bulk densities from trial one, along with texture and organic matter from trial two, were used to compute soil water properties in these soil layers (Saxton and Rawls, 2006) . Data from trial 1 were used to parameterize the 60 90 a nd 90 120 soil layers at trial two. Soil parameters required by APSIM include organic carbon. We used measured values of t otal soil carbon as a proxy for organic carbon. As pH values were below 6 at all agronomic trials (Table 2), we judged our soil s to be low in inorganic calcium carbonates. Furthermore, our measurements of total carbon fall well within the range of organic carbon values found in Daily rainfall data gathered by the extension plannin g area office at each study site from 2011 2013 were used for purposes of model calibration. Temperature and solar radiation data were obtained from the National Centers for Environmental Protection Climate Forecast System Version Two, which includes da ily meteorological data at all points on the globe at a resolution of 0.25 degrees latitude and longitude (Saha et al. , 2014) . Calibration models were run from October 2011 through September 2013, re - setting all soil water, soil C, soil N, and surface orga nic matter parameters at the end of each growing season. The APSIM phosphorus module was not used, and P dynamics were not simulated. Modeled yields from the 2012 - 2013 season were then compared with measured yields for that season. Models were calibrat ed by adjusting non - soil model parameters until the modeled yields fell as close as possible to measured yields (Table 3). Measured yields from the two experimental trials at the Lintipe and Kandeu sites were used in model calibration and validation. 26 At the Golomoti site, models were calibrated using trial 1 only. It should be noted that planting times were adjusted in model calibration, resulting in cowpea being planted after maize in the best calibrated models at each site (Table 3). In contrast, maiz e and cowpea were planted simultaneously in our field trials. At Kandeu and Lintipe, we selected an optimal calibration for each location using a cross - calibration procedure modified from Xiang et al (2008). The calibrated model for trial 1 at each site w as validated against trial two by re - parameterizing it with soil data from trial two and comparing modeled yields with measured yields from trial two. The calibrated model for trial two was validated against trial one in a similar fashion. We identified the model calibration which, when validated, showed the least deviation between modeled and observed yields. This calibration was used in all further simulation exercises. As the soils in the two trials differed considerably in both texture and organic m atter (Appendix A), cross - validation helped ensure that the model was capable of accurately simulating crop growth under a range of soil conditions. Table 3 : Parameters manipulated during model calibration . Values listed are the final values used in lon g - term simulation of cropping systems. Kandeu (trial 2) Lintipe (trial 1) Golomoti (trial 1) Fbiom (fraction organic carbon in microbial biomass; 0 - 15, 15 - 30, 30 - 60, 60 - 90 and 90 - 120 cm) 0.015, 0.015, 0.015, 0.01, 0.01 0.015, 0.015, 0.015, 0.01, 0.01 0.0 1, 0.01, 0.01, 0.01, 0.01 Finert (fraction inert organic carbon; 0 - 15, 15 - 30, 30 - 60, 60 - 90 & 90 - 120 cm) 0.8, 0.8, 0.8, 0.8, 0.9 0.67, 0.67, 0.67. 0.8, 0.9 0.8, 0.8, 0.9, 0.9, 0.9 Maize variety pan6671 r215 pan6671 Maize XF (root penetration factor; 0 - 15 , 15 - 30, 30 - 60, 60 - 90, 90 - 120 cm) 0.7, 0.7, 0.7, 0.4, 0.3 0.5, 0.5, 0.5, 0.2, 0.2 0.4, 0.4, 0.4, 0.3, 0.3 Maize planting density (plants / m 2 ) 4 6 4 Maize topdress (days after planting) 60 72 52 Cowpea variety spreading spreading Spreading Cowpea XF (r oot penetration factor; 0 - 15, 15 - 30, 30 - 60, 60 - 90, 90 - 120 cm) 0.3, 0.3, 0.3, 0.2, 0.2 0.5, 0.5, 0.5, 0.5, 0.5 0.3, 0.3, 0.3, 0.2, 0.2 Cowpea planting density (plants /m 2 ) 5 10 5 Cowpea planting window 01 jan 15 feb 01 jan 15 feb 01 jan 15 feb 27 Ex ploratory modeling We performed exploratory modeling to contrast the performance of monocrop and intercrop systems across 26 growing seasons under varying input conditions and fertility regimes. Models were run from October 1 1979 through September 30 200 5. A complete set of model parameters used in exploratory simulation is included in Appendix B. Temperature and rainfall data for long - term simulations were obtained from Malawi Meteorological Service stations that were located nearby and at similar elev ation to the study sites. The best station to represent each study location was selected by comparing rainfall values measured by the extension planning area office at each study location for the years 2010 2013 with nearby meteorological stations (Figu re 1). Solar radiation data were obtained from the National Centers for Environmental Protection Climate Forecast System reanalysis, which includes daily meteorological data at all points on the globe at a resolution of 0.3 degrees latitude and longitude (Saha et al. , 2010) . 28 Figure 1: Comparison of cumulative precipitation at the study locations with precipitation at nearby meteorological stations. Locations are Kandeu (a), Lintipe (b) and Golomoti (c) while meteorological stations are Chileka (a), Dedza (b) and Monkey Bay (c). Data from study locations are represented by solid lines; data from meteorological stations by dashed lines. 0 200 400 600 800 1000 1200 Oct 2010 Nov 2010 Dec 2010 Jan 2011 Feb 2011 Mar 2011 Apr 2011 May 2011 Jun 2011 Jul 2011 Aug 2011 Sep 2011 Oct 2011 Nov 2011 Dec 2011 Jan 2012 Feb 2012 Mar 2012 Apr 2012 May 2012 Jun 2012 Jul 2012 Aug 2012 Sep 2012 Oct 2012 Nov 2012 Dec 2012 Jan 2013 Feb 2013 Mar 2013 Apr 2013 Cumulative annual precipitation (mm) a) 0 200 400 600 800 1000 1200 Oct 2010 Nov 2010 Dec 2010 Jan 2011 Feb 2011 Mar 2011 Apr 2011 May 2011 Jun 2011 Jul 2011 Aug 2011 Sep 2011 Oct 2011 Nov 2011 Dec 2011 Jan 2012 Feb 2012 Mar 2012 Apr 2012 May 2012 Jun 2012 Jul 2012 Aug 2012 Sep 2012 Oct 2012 Nov 2012 Dec 2012 Jan 2013 Feb 2013 Mar 2013 Apr 2013 Cumulative annual precipitation (mm) b) 0 200 400 600 800 1000 1200 Oct 2010 Nov 2010 Dec 2010 Jan 2011 Feb 2011 Mar 2011 Apr 2011 May 2011 Jun 2011 Jul 2011 Aug 2011 Sep 2011 Oct 2011 Nov 2011 Dec 2011 Jan 2012 Feb 2012 Mar 2012 Apr 2012 May 2012 Jun 2012 Jul 2012 Aug 2012 Sep 2012 Oct 2012 Nov 2012 Dec 2012 Jan 2013 Feb 2013 Mar 2013 Apr 2013 Cumulative annual precipitation (mm) c) 29 Cowpea monocrop, maize monocrop, and maize / cowpea intercrop systems were modeled in long term simulations. Each system was modeled under two nutrient input conditions: 0kg N / ha, and 69kg N / ha split into starter (23kg N / ha) and topdress (46kg N / ha) applications. For Lintipe and Kandeu, model parameters in long - term simulations were taken from the calibr ated model that displayed the least deviation from observed yields when validated. For Golomoti, calibrated parameters for trial one were used. Long - term models were run using two levels of soil organic matter. From soil samples taken from 0 - 15 cm and 15 - with the highest total C and the soil with the lowest total C (Table 4). Using total C and total N values from the high and low C soil, we re - calculated soil water properties and re - pa rameterize the model for organic C, C:N ratio, and soil water properties. All other model parameters remained identical between the high organic matter and the low organic matter soil. Table 4 : Soil C and N used in high organic matter and low organic matt er soils in long - term simulation models. study location (Appendix A). High organic matter soil Low organic matter soil % total C % total N % total C % total N Lintipe 0 - 15 cm 2.241 0 .132 1.291 0.079 15 - 30 cm 1.894 0.089 1.028 0.071 Golomoti 0 - 15 cm 1.250 0.096 0.665 0.060 15 - 30 cm 1.228 0.093 0.653 0.058 Kandeu 0 - 15 cm 0.819 0.070 0.446 0.050 15 - 30 cm 0.760 0.063 0.312 0.043 Exploratory models were first run without croppin g history effects. To accomplish this, we re - set all soil water, soil C, soil N, and surface organic matter parameters at the end of each growing season. This treated each simulated season as though it were the first season of establishment for the syste m. Models where then run a second time without re - sets, allowing 30 effects of cropping history to accumulate. This simulated cropping systems grown for 26 consecutive years. Statistical analysis A nalysis of a gronomic trials Analysis of variance models wer e used to test for treatment effects in all agronomic trials. and unfertilized maize monocrop in all trials, and between fertilized maize in monocrop and ferti lized maize in intercrop with cowpea at Kandeu trials one and two. Stability analysis We performed stability analysis on the total calorie and protein yields from maize cowpea intercrop, and for the two crops in monocrop (as in a field planted half in ma ize and half in cowpea) . Simulated yields in each year were plotted against an environmenal index consisting of the mean yield for all systems in each year . Sensitivity analysis provides a visual representation of how systems perform relative to one anot her under a range of environmental conditions. The approach can be used to assess how stable systems are in the face of inter - annual or geographic variation (Hildebrand and Russell 1996). Stability analysis has been used to explore the stability of maize / cowpea intercrop systems across a range of geographic locations (Faris et al. 1983). Stability analysis has also been used, as in our case, to depict yield stability across a range of growing seasons (Lightfoot et al. , 1987). Risk analysis We performed risk analysis on each cropping system. In risk analysis, systems are compared based on their chance meeting or exceeding a minimal acceptable value (Hildebrand 31 and Russell, 1996) . For smallholder farming households, the chance of producing enough food t o meet household food needs has been used as a minimum acceptable value in risk analysis (Kamanga et al. , 2010; Snapp et al. , 2014) . The calories and protein per - acre that must be produced to meet the nutritional needs of a typical farming household were calculated based on mean household size and landholding at each study location,. We obtained average daily calorie and protein requirements for children 0 14 years of age, adults 15 69 years engaged in moderate to heavy activity, and seniors 70 year s and older, from the Food and Agriculture Organization (2004) and World Health Organization (2007) . We determined the average number of children 14 and under, adults 15 69, and seniors 70 and over, along with average household landholding, based on our survey of households at the Africa RISING study locations (Table 5). Table 5 : Household composition and landholding based on surveys of farming households at the study locations. Calorie and protein production per hectare required to meet household nee Children under 15 Adults 15 69 Seniors 70 and over Landholding (ha) Calories req. (Mcal / ha / yr) Protein req. (kg / ha / yr) Lintipe 2.40 2.65 0.11 0.71 6097 114 Golomoti 2.48 2.54 0.09 0.83 512 1 95 Kandeu 2.21 2.80 0.16 0.89 5005 94 Average 2.36 2.66 0.12 0.81 5365 100 Risk was calculated for each location as the probability of a cropping system meeting or exceeding the protein and calorie requirements of a typical house hold. We calculated calorie and protein production from each cropping system based on nutritional values obtained for crops in Malawi (N. Drost, unpublished data) . The probability of each household meeting its calorie and protein needs with each cr opping system was calculate d as P ht = P ( y cst * a h h c y pst * a h h p ) 32 Where P ht is the probability of household h meeting its calorie and protein needs using treatment t , y cst and y pst are the simulated calorie and protein yield per hectare in season s for treatment t, a h is the area of arable land farmed by household h , in hectares, and r h c and r h p represent the total calorie and protein requirements for household h (Snapp et al. , 2014) . Treatments t included maize cowpea intercrop at 0 kg N / ha, maize and cowpea monocrops at 0 kg N / ha (as in a field split half into maize and half into cowpea) maize cowpea intercrop at 69 kg N / ha, and maize and cowpea monocrops at 69 kg N / ha. 33 RESULTS Farming household characteristics and practices Use of legume crops by farmers differed by study location. At Kandeu groundnut was the most commonly grown legume, and it was not typically grown in intercrop with maize (Figure 2). Common bean, soybean, and cowpea were also relatively commonly grown at Kandeu, and most households growing these legumes intercropp ed them with maize. At Lintipe common bean was by far the most commonly grown legume, and most households growing common bean grew it in intercrop with maize. Groundnut was the second most commonly grown, and very few households intercropped groundnut wi th maize. Few households at Lintipe grew cowpea, and anecdotal reports from farmers and local extensionists suggest that it is particularly prone to attack by aphids at this site , though the causes of this susceptibility are not known . At Golomoti cowpea was the most commonly grown legume, and most households growing cowpea intercropped it with maize. Other legumes were not commonly grown at Golomoti, and anecdotal reports suggest that other legumes are unable to perform well in the dry conditions found at this site. 69% of households reported growing maize in intercrop with legumes, while 41% reported growing maize in monocrop (Figure 3). The majority of farmer s growing maize in intercrop or monocrop applied mineral fertilizer, while a smaller but still considerable portion applied animal manure or vegetable compost to maize. Some farmers used both mineral fertilizer and manure or vegetable compost. Legumes gr own in sole crop were almost never fertilized, either with mineral fertilizer or manure / compost. 34 Figure 2: Percent of households growing legume crops, and percent growing legumes in intercrop with maize. Figure 3: Percent of households growing sol e maize, sole legumes, and maize legume intercrops, and percent applying mineral N fertilizer and manure to these systems. Manure includes animal waste and vegetable compost. Note that some households applied both fertilizer and manure. 0 10 20 30 40 50 60 70 80 90 100 Total Maize intercrop Total Maize intercrop Total Maize intercrop Kandeu Lintipe Golomoti Percent of households growing Cowpea Bean Groundnut Soybean Pigeonpea 0 10 20 30 40 50 60 70 80 90 100 Sole maize Sole legume Maize + legume Percent of households growing Total W/ fertilizer W/ manure 35 Agronomic trials There was a positive effect of fertilization with 69 kg N / ha in all agronomic trials (Table 6; Figure 4). There was no significant effect of intercropping on maize yields at either of the trials at Kandeu. Maize cowpea intercrops were not grown at the Lintipe and Golomoti locations. Table 6 : Comparison of mean yields in agronomic trials at all study locations in the 2012 - 2013 growing season. P - values for comparisons of tria ls tests. Croppin g systems are abbreviated as Mz 0N = unfertilized sole mai ze; Mz 69N = sole maize with 69 kg N / ha and 22 kg P / ha applied; Inter 69N Mz = maize yield in intercrop with co wpea with 69 kg N / ha and 22 kg P / ha applied. Df treat. (residual ) Sum of squares Mean square F Model p - val ue Comparison Comp. p - value Kandeu 1 2(6) 28798195 14399098 70.56 0.0001 Mz 69N / Inter 69N Mz 0.343 3 Mz 0N / Mz 69N 0.0001 Kandeu 2 2(6) 14061350 7030675 55.84 0.0001 Mz 69N / Inter 69N Mz 0.809 8 Mz 0N / Mz 69N 0.0002 Lintipe 1 1(4) 4008 0576 40080576 236 0.0001 Mz 0N / Mz 69N 0.0001 Lintipe 2 1(4) 36469791 36469791 542.6 0.0001 Mz 0N / Mz 69N 0.0001 Golomoti 1 1(4) 8368305 8368305 93.13 0.0006 Mz 0N / Mz 69N 0.0006 Model calibration and validation After model calibration, crop simulat ions were able to reasonably approximate observed yields in the 2012 - 2013 season. In most cases, simulated yields were brought within one standard error of the mean observed yield after calibration (Figure 4). Model calibrations for some study locatio ns performed better in crossvalidation than others. When model parameters calibrated for Kandeu 1 were used to simulate Kandeu 2, maize at 0 kg N / ha produced nearly double the observed yields. When the Kandeu 2 calibration was used to simulate Kandeu 1 , most simulated yields reasonably approximated observed yields, with maize at 69 kg N / ha 36 noticeably under - yielding. When the Lintipe 1 calibration was used to simulate Lintipe 2, simulated yields fell very close to observed yields. However, when the L intipe 2 calibration was used to simulate Lintipe 1, maize under - yielded considerably in monocrops at 69 kg N / ha. We used calibrated models for Kandeu 2, Lintipe 1, and Golomoti 1 in all further simulations. Figure 4 : Calibration and crossvalid ation of crop simulation models. Shaded bars are mean simulated grain yields with error bars showing one standard error of the mean . White bars are modeled yields. Croppin g systems are abbreviated as Mz 0N = sole mai ze with 0kg N / ha applied; Mz 69N = s ole maize with 69 kg N / ha applied; Cp 12N = sole cowpea with 12 kg N / ha applied; Inter 69N Mz = maize yield in intercrop with cowpea with 69 kg N / ha applied; Inter 69N Cp = cowpea yield in intercrop with maize with 69 kg N / ha applied. 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp a) Kandeu 1 calibration 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp b) Kandeu 1:2 crossvalidation 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp c) Kandeu 2 calibration 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp d) Kandeu 2:1 crossvalidation 37 Figure 4 (c 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp e) Lintipe 1 calibration 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp f) Lintipe 1:2 crossvalidation 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp g) Lintipe 2 calibration 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp h) Lintipe 2:1 crossvalidation 38 Modeled yields In long - term simulations (1980 2005) in which soil parameters were re - set every year to eliminate effects of cropping history, the relationship between maize yields in intercrop and yields in monocrop varied by study location and by rates of N input (Figure 5 , Figure 6 ). When 0 kg N / ha were applied, intercropped maize yields at the Lintipe study location under all soil conditions were approximately half of maize yields in comparable monocrops. In co ntrast, at the Kandeu and Golomoti study locations, maize yields in intercrop and monocrop were equivalent under all soil conditions when 0 kg N / ha were applied. When 69 kg N / ha were applied, maize yields in intercrop were roughly equivalent to yields in monocrop at all study locations and in all soils. Cowpea yields in intercrop were lower than yields in monocrop at all study locations, soils and N rates. 0 1000 2000 3000 4000 5000 6000 Mz 0N Mz 69N Cp 12N Inter 69N Mz Inter 69N Cp i) Golomoti 1 calibration 39 Figure 5 : Modeled grain yields of cropping systems from 1980 2005 . Soil parameters wer e re - set annually to remove effects of cropping history at the Kandeu (a), Lintipe (b), and Golomoti (c) study location s. ems a bbreviated as Mz mono = maize monocrop, Mz inter = maize intercropped with cowpea, Cp mono = cowpea monocrop, Cp inter = cowpea intercropped with maize. Bars show mean yield across all years; error bars show one standard error of the mean. 0 1000 2000 3000 4000 5000 6000 Grain yield (kg / ha) a) Mz mono Mz inter Cp mono Cp inter 0 1000 2000 3000 4000 5000 6000 Grain yield (kg / ha) b) 0 1000 2000 3000 4000 5000 6000 Low OM 0N Low OM 69N High OM 0N High OM 69N Grain yield (kg / ha) c) 40 Figure 6 : Modeled biomass yields of cropping systems from 1980 2005. S oil parameters were re - set annually to remove effects of cropping history at the Kandeu (a), Lintipe (b), and Golomoti (c) study location s. ems abbreviated as Mz mono = maize monocrop, Mz inter = maize intercropped with cowpea, Cp mono = cowpea monocrop, Cp inter = cowpea intercropped with maize. Bars show mean yield across all years; error bars show one standard error of the mean. 0 2000 4000 6000 8000 10000 12000 Biomass yield (kg / ha) a) Mz mono Mz inter Cp mono Cp inter 0 2000 4000 6000 8000 10000 12000 Biomass yield (kg / ha) b) 0 2000 4000 6000 8000 10000 12000 Low OM 0N Low OM 69N High OM 0N High OM 69N Biomass yield (kg / ha) c) 41 When soil parameters were not re - set every year, allowing yields to be affected by cropping history, intercropped maize tended to out - yield equivalent maize monocrops (Appendix C Figure 1) . This was particularly true under unfertilized conditions, where intercropped maize drastically out - performed maize monocrop at all study locations and in all soils. When 69 kg N / ha were applied, maize in intercrop produced somewhat higher yields tha n maize in monocrop under all soil conditions at the Kandeu location, while monocropped and intercropped maize produced equivalent yields at Lintipe and Golomoti. Cowpea produced lower yields in intercrop than in monocrop at all study locations and under all soil conditions. We ran models in which soil parameters were not re - set as an exploratory exercise, to see whether the relative advantages of intercrop and monocrop systems might differ due to the effects of cropping history. In APSIM, cropping histo ry effects are primarily driven by biomass production and its contribution to soil C and N reserves (Probert et al. , 1998) . Our confidence in these models is therefore limited, as we did not calibrate them based on measures of biomass production in the fi eld. These results are the product of built - in assumptions in APSIM about the ratio of grain yield to biomass production for the cultivars used in our simulations, and the accuracy of these assumptions in our study locations has not been supported with d ata from the field. We therefore chose to perform all further analysis using models in which soil parameters were re - set every year. Stability analysis The relationship of calorie yields in maize cowpea intercrops to equivalent maize and cowpea monocrops varied by study locatio n and N input (Table 7, Figure 7 ). In Kandeu, calorie yields in intercrop were higher than calorie yields in monocrop across the full range of environmental conditions, from the most favorable season and soil to the least favorable season 42 and soil. This was true under both 0 kg N / ha and 69 kg N / ha. At Lintipe, on the other hand, calorie yields in intercrop were lower than those in monocrop under the full range of environmental conditions when 0 kg N / ha were applied, and highe r when 69 kg N / ha were applied. At Golomoti, calorie yields were higher in intercrop at 69 kg N / ha under the full range of environmental conditions. When 0 kg N ha were applied, calorie yields were similar in monocrop and intercrop under less favorab le environmental conditions, and higher in intercrop under more favorable conditions. The relationship of protein production in maize cowpea intercrops to protein production in equivalent monocrops also varied by s tudy location (Table 7, Figure 7 ). In Kan deu, intercrops consistently produced more protein / ha than monocrops when fertilized with 69 kg N / ha. With no N fertilization, intercrops produced similar amounts of protein to monocrops when environmental conditions were unfavorable, and more than mo nocrops when environmental conditions were favorable. In Lintipe, intercrops produced less protein than monocrops under the full range of environmental conditions, regardless of N fertilization. At Golomoti, intercrops and monocrops fertilized with 69 kg N / ha produced similar amounts of protein in unfavorable environmental conditions. Under more favorable conditions, intercrops began to produce more protein than monocrops. In unfertilized cropping systems at Golomoti, intercrops produced less protein than monocrops under unfavorable environmental conditions, and more than monocrops under favorable environmental conditions. Calorie and protein production in maize / cowpea intercrop systems tended to be less stable across environmental conditions than in equivalent m onocrops (Figure 7 ). This can be observed in the steeper slopes of intercrop yields when plotted against the environmental index, indicating that intercrops exhibit a sharper response to changes in environmental conditions. In 43 many cases, th is was due to intercrops producing higher calorie yields than monocrops under the most favorable environmental conditions, and equivalent or only somewhat higher yields under unfavorable conditions. Therefore, though intercrop systems were less stable tha n monocrops, they tended to produce more calories and protein than monocrops under the full range of environmental conditions. Exceptions to this general trend occurred in Lintipe. When 0 kg N / ha were applied at Lintipe, the intercrop system produced v ery low calorie yields under less favorable conditions, and only approached calorie yields in monocrop under more favorable conditions. Protein yields were consistently lower in intercrop than in monocrop at Lintipe. Protein yield stability was similar i n monocrop and intercrop systems. 44 Figure 7 : Stability analysis of calorie and protein yields from modeled cropping systems . Given as Kandeu calories (a), Kandeu protein (b), Lintipe calories (c), Lintipe protein (d), Golomoti calories (e), Gol omoti protein (f). Calorie yields for each system in each year are plotted against an environmenal index consisting of the mean for all systems in each year. Datapoints were fit using linear regressions (Table 7) . Cropping systems are abbreviated as Mon o 0N = maize and cowpea monocrop at 0 kg N / ha; Mono 69N = monocrop at 69 kg N / ha; Inter 0N = maize cowpea intercrop at 0 kg N / ha; Inter 69N = intercrop at 69 kg N / ha. 0 5000 10000 15000 20000 4000 6000 8000 10000 Calorie yield (Mcal / ha) a) 0 100 200 300 400 500 150 250 350 Protein yield (kg / ha) b) EI Mono 0N Mono 69N Inter 0N Inter 69N 0 5000 10000 15000 20000 6000 8000 10000 12000 Calorie yield (Mcal / ha) c) 0 100 200 300 400 500 150 250 350 Protein yield (kg / ha) d) 0 5000 10000 15000 20000 2000 4000 6000 8000 Calorie yield (Mcal / ha) Environmental Index (EI) e) 0 100 200 300 400 500 0 100 200 Protein yield (kg / ha) Environmental Index (EI) f) 45 Table 7 : R 2 values for linear regressions used in stability analysis plots . Stability analysis plots are given in Figure 7 . Mono 0N Mono 69N Inter 0N Inter 69N Kandeu Calories 0.727 0.818 0.512 0.869 Protein 0.740 0.687 0.699 0.787 Lintipe Calories 0.682 0.801 0.655 0.806 Protein 0.800 0.855 0.616 0.594 Golomoti Calori es 0.715 0.732 0.722 0.907 Protein 0.673 0.651 0.711 0.832 Risk analysis The chance of a typical household meeting their calorie and protein needs with their own production was influenced by location, N input, soil organic matte r, and cropping system ( Figure 8 ). Households at Lintipe had the highest overall chance of meeting calorie and protein needs, followed closely by Kandeu. At Golomoti, the chance of meeting calorie and protein needs was much lower. At all sites, households had a higher chance o f meeting their calorie and protein needs when 69 kg N / ha were applied than under unfertilized conditions. At the Kandeu and Lintipe study locations, households met 100% of their needs under fertilized conditions in every year that the simulation was ru n . At all sites, high organic matter soils produced a higher chance of meeting calorie and protein needs than low organic matter soils. However, this effect was slight in comparison with the effect of soil fertility. The typical household experienced a higher chance of meeting household calorie and protein needs with their own production using maize cowpea intercrop than using maize and cowpea monocrops at the Kandeu and Golomoti study locations. This intercrop advantage was present both in unfertilized and fertilized systems, and in low organic matter and high organic matter soils. At Lintipe, on the other hand, households experienced a higher chance of meeting 46 calorie and protein needs using unfertilized maize and cowpea monocrops than using an unfert ilized maize / cowpea intercrop. The reverse was true under fertilized conditions. 47 Figure 8 : Probability that average households will meet calorie and protein needs with their own production. Given for low organic matter soil at Kandeu (a), high organic matter soil at Kandeu (b), low organic matter soil at Lintipe (c) , high organic matter soil at Lintipe (d), low organic matter soil at Golomoti (e) and high organic matter soil at Golomoti (f). The chance that a cropping system fails to meet hous ehold calorie and protein needs is plotted against percent of household calorie and protein needs. Therefore, a point at 100% on the y axis and 40% on the x axis indicates that a system fails to meet 100% of household calorie and protein needs 40% of the time . 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs a) 0 50 100 150 200 250 300 350 0 25 50 75 100 b) Mono 0N Inter 0N Mono 69N Inter 69N 100% needs 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs c) 0 50 100 150 200 250 300 350 0 25 50 75 100 d) 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs % chance cal and prot yield < needs e) 0 50 100 150 200 250 300 350 0 25 50 75 100 % chance cal and prot yield < needs f) 48 DISCUSSION In all study locations and cropping systems examined, maize was severely N limited in unfertilized systems. Addition of 69 kg N / ha, the current recommended rate for maize in Malawi, increased maize yields in intercrop and monocrop drama tically. Fertilized maize produced the highest yields at Lintipe, followed by Kandeu, while producing lower yields at Golomoti (Figures 5 and 6). This trend in N response is very likely due to water limitation at Golomoti and Kandeu. While the three sit es do not differ greatly in annual rainfall (Figure 1), mean temperatures are lowest at Lintipe, higher at Kandeu, and highest at Golomoti. This would increase the rate of evapotranspiration, depleting soil reserves at the latter sites. In addition, soil s at Lintipe have higher clay content than Kandeu or Golomoti (Table 2), making for increased water holding capacity at Lintipe. Decreased N response under water - limited conditions has frequently been observed in both maize monocrop and maize intercrop sy stems (Kurtz et al. , 1952; Yamoah et al. , 2000) , supporting water limitation as a cause for the differing N response between sites. Cowpea did not appear to be N limited in any of the cropping systems examined in this study. Cowpea yields in monocrop rem ained relatively constant across 0 kg N / ha and 69 kg N / ha treatments (Figures 5 and 6). Cowpea typically does not exhibit a strong response to soil applications of N (Elowad and Hall, 1987) , which is to be expected given its ability to supply the majo rity of N needs through biological N fixation (Vesterager et al. , 2008) , though response to N fertilization has been observed in some cowpea systems (Rusinamhodzi et al. , 2012) . The lack of N response by cowpea also accords with farmer practice. Farmers in our study locations virtually never applied fertility inputs to sole legume crops (Figure 2). This suggests that farmers 49 are fully aware that they will achieve a greater response by applying their limited fertility resources to maize - based systems. Rel ative yields of maize and cowpea in monocrop and intercrop differed between sites when each season was treated as the first season of establishment. Under unfertilized conditions, sole cropped maize produced comparable grain yields in monocrop and intercr op at both Kandeu and Golomoti, while cowpea yield was suppressed in intercrop (Figure 5). At Lintipe, on the other hand, both maize and cowpea yields were suppressed in intercrop under unfertilized conditions. This indicates that maize was competitively dominant under unfertilized conditions at Kandeu and Golomoti, while at Lintipe neither maize nor cowpea were clearly dominant, resulting in the competitive suppression of both crops. In the few studies examining competition in unfertilized, additive mai ze cowpea monocrops, yields of both crops have been found to be un - altered (Rusinamhodzi et al. , 2012) , or cowpea has been suppressed by maize (Ofori and Stern, 1987) . The latter case supports our results at Kandeu and Golomoti, while in the former case, use of a lower maize planting density than was used in our study may have reduced the competitive effects of maize on cowpea. The suppression of maize by cowpea in unfertilized intercrops at Lintipe, but not at other study locations, may have been caused by the high yield potential of cowpea at Lintipe. Cowpea monocrop produced higher grain yields at Lintipe than at other sites, both in our agronomic trials (Figure 4) and in our models (Figure 5). The more vigorous growth of cowpea at Lintipe suggests t hat it was able to exert a greater degree of competitive pressure on its intercrop partner. When N fertilizer was added to the system, maize became dominant at all sites. Fertilized maize yields were not suppressed in intercrop at any of our study loca tions. At the same time, cowpea yields were suppressed more strongly in intercrop under fertilized conditions 50 than under unfertilized conditions, likely due to the more vigorous growth of maize. In studies of fertilized maize cowpea intercrop systems, co mpetitive dynamics have varied. Both maize and cowpea yields can be suppressed in fertilized intercrop systems (Ofori and Stern, 1987; Watiki et al. , 1993) , though this suppression may occur only in seasons with unfavorable rainfall and not in more favora ble seasons (Hulugalle and Lal, 1986) . In other systems (Rusinamhodzi et al. , 2012) , as in our own agronomic trials (Figure 4, Table 6), maize yields have not been suppressed in fertilized intercrop systems. Our results suggest the need for further inves tigations to determine the environmental conditions under which cowpea will act as a strong competitor with maize. When we included the effects of cropping history in our models, the competitive dynamics of maize cowpea intercrop systems were altered. All owing biomass produced by previous crops to affect the growth of subsequent crops caused maize yields in unfertilized conditions to became higher in intercrop than in monocrop (Appendix C Figure 1). Under fertilized conditions, however, this improvement i n maize yield due to intercropping was dampened or eliminated. Improved maize yields in intercrops under unfertilized conditions were almost certainly due to the provision of N to maize through the deposition of N - rich cowpea biomass in previous seasons ( Vesterager et al. , 2008) . Other modeling work using APSIM in Malawi in which soil parameters were not re - set has shown improved maize yield and yield stability in maize pigeonpea additive intercrop systems in successive years of establishment (Ollenberger , 2012) . We cannot have complete confidence that these results reflect the behavior of actual cropping systems in Malawi, as we did not validate our models for biomass production. However, our results do suggest that further investigation of the long - ter m effects of maize cowpea intercropping on soil fertility and maize yields may be warranted. 51 Maize cowpea intercrop systems were typically less stable than monocrops of maize and cowpea at Kandeu and Golomoti. At both sites, calorie and protein yields fro m maize cowpea intercrop systems exhibited a stronger response to variation in environmental conditions than did the two monocrops (Figure 7 ). The same was true for calorie yields at Lintipe, though the stability of protein yields remained constant in m onocrop and intercrop systems. The reduction in stability under intercrop systems often went hand in hand with increased yields, however. At Kandeu, calorie yields from intercrop systems across the full range of seasons and soils far exceeded yields from monocrops when 69 kg N / ha were applied. At 0 kg N / ha the intercrop advantage was less pronounced, but still apparent. Protein yields in intercrop and monocrop at Kandeu were similar under less favorable conditions, but intercrops experienced an adva ntage under more favorable conditions. A similar pattern was observed at Golomoti for both calorie and protein yields. At this site, intercrop yields under less favorable environmental conditions were similar to yields in monocrop, but under more favorab le conditions intercrops always displayed an advantage. The lower yield stabilities observed at Kandeu and Golomoti were thus due to intercrops exhibiting a more robust response to favorable conditions than monocrops. This illustrates that yield stabilit y alone does not provide enough information to discern whether a cropping system is likely to be advantageous to farmers. At Lintipe, calorie production was likewise less stable in intercrops than in monocrops. While systems fertilized with 69 kg N / ha f ollowed the same pattern observed at Kandeu, calorie yields in unfertilized intercrop systems were lower than monocrops across the full range of environmental conditions. This resulted from the competitive suppression of maize by cowpea in unfertilized in tercrop systems at Lintipe. Maize contains more calories / kg than cowpea, and suppression of maize therefore led to a sharp decline in calorie yields. Protein yields were 52 consistently higher in monocrop than in intercrop at Lintipe, with monocrops and i ntercrops demonstrating similar yield stability. This was likely due to the superior cowpea yields in monocrop at Lintipe. Cowpea grain is much more protein rich than maize. Therefore, since cowpea yields were sharply suppressed in intercrop, monocrops of maize and cowpea were able to produce more protein than intercrops at the site where sole cowpea performed best. We hypothesized that maize cowpea intercrop systems would produce more calories and protein than monocrops under unfertilized conditions, bu t not under fertilized conditions. In fact, intercrops tended to produce higher calorie and protein yields than monocrops under both fertilized and unfertilized conditions. Our hypothesis was largely based on studies that found maize yields to be equival ent in additive monocrop and intercrop systems under unfertilized conditions, but depressed in intercrop with cowpea under fertilized conditions (Ofori and Stern, 1987; Watiki et al. , 1993) , though in one case this effect was found only in a dry season and not in a wet season (Hulugalle and Lal, 1986) . In our own agronomic trials (Table 1, Figure 4), as in at least one published case spanning multiple seasons (Rusinamhodzi et al. , 2012) , maize yield was not suppressed in intercrop with cowpea under fertili zed conditions. We are aware of only one experimental study contrasting the total productivity of maize cowpea intercrops and maize and cowpea monocrops across varying fertility levels (Ofori and Stern, 1987) . In this case, reduction in the LER with inc reasing N fertilization was cause by suppression of maize in intercrop with cowpea in N fertilized systems. In our own case, because maize was not suppressed in competition with cowpea, the addition of a cowpea harvest in intercrop caused fertilized inter crop systems to achieve calorie and protein yields that were consistently higher than monocrops. As we have seen in our own study and others, competitive dynamics between maize and cowpea can vary by location (Ofori and Stern, 1987; 53 Rusinamhodzi et al. , 2 012) . Our results therefore suggest the need for further experimentation to determine the effects of N fertilization on intercrop advantage in maize cowpea systems. Such experimentation should be carried out over multiple cropping seasons, as there is e vidence to suggest that competitive dynamics between maize and cowpea can change from one season to the next (Hulugalle and Lal, 1986) . Our second hypothesis, that maize cowpea intercrop systems would be less stable than maize and cowpea monocrops, was lar gely c onformed by this study (Figure 7 ). However, as previously noted, in most cases this was due to greater intercrop advantage under favorable environmental conditions. Reduced stability of this kind may actually be favorable from a e, as intercropping can allow them to achieve higher yields in the best seasons, and produce yields at least as high as monocrop in the worst seasons. Intercropping decreased risk to farmers in most situations, improving the average eeting cal orie and protein needs (Figure 8 ). Fertilized maize cowpea intercrop systems had a higher chance of meeting household calorie and protein needs than maize and cowpea monocrops at all study locations. Under unfertilized conditions, intercrops ha d a higher chance of meeting household needs at Kandeu and Golomoti, but not at Lintipe. As previously mentioned, maize grain yields were suppressed in competition with cowpea in unfertilized intercrops at Lintipe (Figure 5). This caused calorie and prot ein yields to be lower in this system across the full range of en vironmental conditions (Figure 7 ). Consequently, unfertilized intercrops were less likely to meet household calorie and protein needs than monocrops at Lintipe. Our general finding that inte rcropping decreases production risk is in agreement with other studies of risk in cereal legume intercrop systems. In Botswana, sorghum cowpea 54 intercrops had a higher chance of meeting a minimum economic yield than sorghum and cowpea monocrops (Lightfoot et al. , 1987) . In Malawi, unfertilized maize pigeonpea intercrops had a higher chance of meeting household maize needs than unfertilized maize monocrops (Kamanga et al. , 2010) . In this system, intercropping maize with a legume produced a reduction in ris k equivalent to adding 35 kg N / ha to sole maize. Our findings also accord partially with a modeling study that found improved economic yields in maize cowpea intercrop compared with maize and cowpea monocrops under unfertilized conditions, but not under fertilized conditions (Carberry et al. , 1996) . As previously mentioned, the relative performance of maize cowpea intercrops and monocrops under fertilized conditions may depend on whether maize yields are suppressed by cowpea, a characteristic which can vary by location (Ofori and Stern, 1987; Rusinamhodzi et al. , 2012) . However, as Carberry et al. (1996) do not include information about the relative performance of maize and cowpea in their models, we cannot pinpoint the cause of the differing results b etween their study and ours. Our findings that maize cowpea intercropping can increase calorie and protein yields and reduce risk for smallholder farmers are supported by farmer practice. Intercropping was a common practice at all sites (Figure 2), with 69% of farmers reporting that they used some form of maize legume intercropping (Figure 3). The majority of farmers who grew cowpea reported growing it in intercrop with maize (Figure 2). This suggests that, based on their own knowledge and experience o f their cropping systems, farmers in our study locations have found that growing cowpea in intercrop with maize is a worthwhile practice. Our results do conflict with farmer practice in one regard. In our agronomic trials (Figure 4) and modeling work (Fig ure 5), we found that cowpea performed best at the Lintipe study location. In our survey of farmer practice, however, we found that cowpea is less commonly 55 grown at Lintipe than at the other study locations. Anecdotal reports from farmers and extensionis ts suggest that cowpea experiences very high pressure from aphids at Lintipe when insecticides are not used. While we did use insecticide in our experimental trials, farmers rarely report using insecticides in the field, often citing cost as a barrier. Th growing conditions. We did not include P dynamics in our simulation models, and all agronomic trials containing cowpea were amended with P fertilizer. While agricultur al systems are not typically P limited in Malawi, there is high geographical variability in extractable soil P (Snapp, 1998) , resulting in scattered patches that are P deficient. We observed this P variability at our ields contained very low Bray extractable P, while other fields showed much higher values (Appendix A). In P deficient soils, in the absence of P inputs, cowpea yields and biological N fixation can be limited (Vesterager et al. , 2008; Rusinamhodzi et al. , 2012) . Farmers in our study location typically apply at least some mineral fertilizer to maize legume intercrop systems (Figure 2). If farmers are able to include some P in their fertilizer applications to maize cowpea intercrops, cowpea performance in intercrop is unlikely to be P limited. However, in cowpea monocrop systems which are rarely fertilized (Figure 3), some In many ways, the intercrop advantage demonstrated by our study is based on conser vative assumptions. By re - setting soil parameters every year, we did not include one of the chief mechanisms through which intercropping systems can achieve superior productivity to monocrops namely the production of N rich legume biomass that can contr ibute N to subsequent cereal crops (Fukai and Trenbath, 1993; Chikowo et al. , 2004) . When models were run without re - setting soil parameters, even greater intercrop advantage was observed, 56 particularly under un - fertilized conditions. Furthermore, cowpea grown in intercrop with fertilized maize (the typical practice in our study locations) may experience an advantage over unfertilized cowpea due to freedom from P limitation a factor which was not considered in our study. On the other hand, our use of in secticides to control aphids and other pests may have resulted in higher cowpea yields than farmers could expect without insecticides, particularly at Lintipe. Despite this drawback, we believe that our results are unlikely to over - state the advantages of maize cowpea intercrops given the conservatism of our other assumptions. We look to further experimentation with maize cowpea intercrops under farmers conditions, as part of the Africa RISING project and others, to complement our simulation results with data from the field. Ultimately, this study will prove useful only to the extent that farmers and extensionists perceive the work to be trustworthy and relevant to their actual systems (Cash et al. , 2003) . They are only likely to do so if they are include d in the research process. We have made the first steps in this direction by including farmers and extensionists in the implementation of our study, and by basing experimental systems on local practice to the extent possible. By sharing our results with farmers and taking their suggestions into account as we design the next phases of research, we can integrate this study into a cycle of iterative co - learning that will increase its legitimacy for farmers and improve the quality of the science (McDougall an d Braun, 2003) . 57 CONCLUSIONS Maize cowpea intercrop systems in central Malawi can produce more calories and protein than maize and cowpea monocrops under most growing conditions and in most seasons. While intercrop systems tend to be less stable than mon ocrops, this is due to their superior performance in seasons that are particularly favorable. Therefore, intercropping maize with cowpea rather their food needs thr ough their own production. The advantages of intercropping are most pronounced and universal in systems where moderate levels of fertilizer are applied. In unfertilized systems, intercrop advantage appears to be location specific. Therefore, while our results suggest that maize cowpea intercropping can be an advantageous practice for smallholder farmers in Malawi, they also point to the need for further research into the conditions under which maize cowpea intercrops perform most favorably. 58 APPENDI CES 59 APPENDIX A Description of soils at the Africa RISING Malawi 2013 intervention sites, including the three study locations used in this wo rk. Soils were sampled in the summer of 2013 (see Materials and Methods above). Part A describes soils s ampled from Africa RISING Table A. 1: Description of soils from Africa RISING mother trials . Site Mother trial Lat / Long Sample type Depth (cm) % Sand % Silt % Clay BD (g/cm 3 ) % Tot C % Tot N pH Bray P (ppm) Lintipe 1: Mbidzi 34.09781 E / 14.20679 S Pit 0 15 10 16 74 1.4 2.6 0.13 5.0 14.5 15 30 14 21 65 1.5 2.3 0.12 4.9 4.5 30 60 25 25 50 1.3 1.1 0.07 4.8 1.0 60 90 6 16 78 1.4 0.7 0.05 5.1 2.5 90 120 5 19 76 1.3 0.4 0.04 5.2 2.5 Hoe 0 15 22 29 48 NA 2.3 0.13 4.9 13.5 15 30 19 26 55 NA 2.2 0.12 4.9 12.5 30 60 18 24 58 NA 2.0 0.11 5.0 8.5 Lintipe 2: Mkuwazi 34.11219 E / 14.19522 S Hoe 0 15 52 17 31 NA 1 .5 0.10 5.1 9.5 15 30 38 16 46 NA 1.7 0.10 5.2 3.0 Golomoti 1: Msamala 34.60472 E / 14.43828 S Pit 0 15 45 23 32 1.4 0.9 0.07 4.9 35 15 30 44 21 36 1.4 0.6 0.05 4.9 22 30 60 31 21 48 1.3 0.2 0.05 4.7 13.5 60 90 27 22 51 1.5 0.2 0. 05 4.3 16.5 90 120 29 21 50 1.7 0.2 0.02 4.2 12 Hoe 0 15 52 22 26 NA 0.9 0.06 4.5 56.5 15 30 43 20 37 NA 0.5 0.05 4.2 20 30 60 34 19 47 NA 0.1 0.04 4.9 12 60 Table A. 1 Site Mother trial Latitude / longitude Sample type Depth (c m) % Sand % Silt % Clay BD (g/cm 3 ) % Tot C % Tot N pH Bray P (ppm) Golomoti 2: Kalumo 34.59544 E / 14.43711 S Hoe 0 15 44 23 32 NA 1.2 0.09 5.7 26.0 15 30 40 22 38 NA 0.6 0.05 5.5 16.5 30 60 36 21 43 NA 0.5 0.05 5.5 11.0 Kandeu 1: Katsese 34 .59762 E / 14.6297 S Pit 0 15 70 18 13 1.5 1.1 0.09 5.0 64 15 30 64 18 18 1.4 1.0 0.08 5.1 37.5 30 60 67 15 18 1.3 0.6 0.05 5.2 14.5 60 90 40 20 40 1.4 1.5 0.12 4.9 9.5 Hoe 0 15 75 17 9 NA 1.0 0.09 Insuf* 72.5 15 30 74 13 13 NA 0 .7 0.06 5.4 42.5 30 60 68 17 16 NA 0.8 0.06 Insuf* 27.5 Kandeu 2: Dauka 34.60672 E / 14.64078 S Hoe 0 15 59 12 29 NA 0.7 0.06 5.0 4.0 15 30 50 10 40 NA 0.4 0.05 4.8 2.0 30 60 47 11 43 NA 0.2 0.04 4.9 2.0 Nsipe 1: Amosi 34.74677 E / 14.92927 S Pit 0 15 49 22 30 1.4 1.3 0.09 4.4 40 15 30 35 26 39 1.5 1.6 0.11 4.5 10 30 60 20 18 62 1.5 1.3 0.10 5.1 4 60 90 16 14 70 1.5 0.7 0.08 4.7 2.5 90 120 25 15 60 1.5 0.4 0.13 4.9 3.5 Hoe 0 15 54 19 28 NA 1.1 0.08 4.7 18.5 15 30 40 17 43 NA 1.5 0.09 4.7 12.5 30 60 24 18 59 NA 1.1 0.09 5.0 5.5 Nsipe 2: Nzililongwe 34.743611 E / 14.907222 S Hoe 0 15 40 21 38 NA 1.6 0.11 5.0 6.0 15 30 29 21 51 NA 1.5 0.11 5.1 5.5 30 60 20 19 61 NA 1.5 0.13 5.2 2.0 NA in s that we were unable to run a test because there was insufficient soil remaining from a given sample . 61 Table A. 2: ntion sites. Site (Village) Fertility ranking Field Latitude / longitude Crop Depth (cm) % Sand % Silt % Clay % Tot C % Tot N pH Bray P (ppm) Lintipe (Mbidzi) High 8 34.084 0 E / 14.2117 S Mz + Bn 0 15 52 22 26 1.4 0.10 5.8 20 15 30 49 25 26 1 .0 0.07 Insuf* 8 9 34.1003 E / 14.2115 S Mz 0 15 51 23 26 1.8 0.12 6.2 26 15 30 56 22 22 1.0 0.07 6.3 10 10 34.1008 E / 14.2033 S Mz 0 15 21 28 51 2.2 0.13 Insuf* 23.5 15 30 24 22 54 1.9 0.09 Insuf* 15.5 Lintipe (Mbidzi) Low 2 34.095 4 E / 14.2066 S Gn 0 15 38 20 42 1.3 0.08 5.7 2 15 30 28 20 52 1.0 0.07 5.9 1 5 34.0925 E / 14.2057 S Sy 0 15 34 21 45 1.6 0.10 5.6 3 15 30 27 22 51 1.4 0.09 5.7 2.5 7 34.0934 E / 14.2045 S Sy + Gn 0 15 28 21 50 1.9 0.11 5.6 3 15 30 17 19 64 1.6 0.09 5.8 2 Golomoti (Msamala) High 3 34.5910 E / 14.4335 S Mz + Pp 0 15 66 15 19 0.9 0.07 6.5 160 15 30 56 17 27 0.7 0.06 Insuf* 160 4 34.5852 E / 14.4392 S Mz 0 15 56 24 19 1.2 0.10 Insuf* 220 15 30 53 26 21 1.2 0.09 I nsuf* 201.5 5 34.5907 E / 14.4397 S Mz 0 15 68 15 16 0.6 0.05 6.3 85 15 30 55 20 25 0.7 0.05 6.2 88 Golomoti (Msamala) Low 1 34.5868 E / 14.4307 S Mz 0 15 63 24 13 0.9 0.08 6.6 77.5 15 30 67 20 13 0.8 0.06 6.8 73.5 8 34.6067 E / 14. 451 8 S Cp 0 15 67 17 17 1.0 0.07 6.3 73.5 15 30 63 18 19 0.9 0.07 6.0 71 9 34.5992 E / 14.4412 S Mz 0 15 62 22 16 0.7 0.06 6.0 36.5 15 30 51 22 27 0.7 0.06 5.1 21.5 62 Table A. 2 Site (Village) Fertility ranking Field Latitude / l ongitude Crop Depth (cm) % Sand % Silt % Clay % Tot C % Tot N pH Bray P (ppm) Kandeu (Katsese) High 3 34.5929 E / 14.6281 S Mz 0 15 75 10 15 0.4 0.05 6 55.0 15 30 69 10 21 0.3 0.04 6 29.0 4 34.5928 E / 14.6286 S Mz 0 15 65 15 19 0.8 0.07 5.8 71.0 15 30 65 12 23 0.8 0.06 5.8 68.0 9 34.5918 E / 14.6266 S Mz 0 15 68 14 18 0.7 0.07 5.7 43.5 15 30 67 13 21 0.5 0.06 5.9 18.0 Kandeu (Katsese) Low 1 34.5973 E / 14.6248 S Mz + Sy 0 15 72 5 23 0.5 0.06 5.5 9.0 15 30 58 10 31 0. 6 0.06 6.4 4.0 2 34.5989 E / 14.6222 S Gn 0 15 71 11 18 0.7 0.07 5.3 4.0 15 30 66 8 26 0.6 0.06 4.7 3.0 6 34.5978 E / 14.617 7 S Mz + Pp 0 15 60 14 26 0.7 0.07 5.2 21.0 15 30 65 8 27 0.6 0.07 5.5 19.0 Nsipe (Amosi) High 4 34.7437 E / 14.9306 S Mz 0 15 65 17 18 0.7 0.07 6.0 104 15 30 57 13 29 0.7 0.07 6.2 93.5 5 34.7454 E / 14.9296 S Mz 0 15 48 17 35 1.2 0.11 5.9 15 15 30 40 16 44 1.2 0.10 5.6 8.5 7 34.7488 E / 14.9333 S Mz 0 15 44 21 34 1.5 0.11 5.8 13.5 15 30 32 23 44 1.1 0.09 Insuf* 12 Nsipe (Amosi) Low 1 34.7454 E / 14.9269 S Mz + Cp 0 15 35 19 47 1.4 0.11 6.0 4.0 15 30 23 20 57 1.4 0.11 6.0 2.5 2 34.7449 E / 14.9263 S Mz 0 15 30 20 50 1.5 0.12 6.1 8.0 15 30 22 21 58 1.6 0.11 6.0 4.5 9 34.7421 E / 14.9325 S Mz + Gn 0 15 63 15 22 0.6 0.06 5.7 8.0 15 30 65 1 34 0.5 0.05 5.7 3.5 Crop abbreviations used: Mz = maize, Bn = bean, Gn = groundnut, Sy = soybean, Pp = pigeonpea, Cp = cowpea that we were unable to run a test because there was insufficient soil remaining from a given sample . 63 APPENDIX B All parameters used in exploratory crop simulation models, run in APSIM 7.5. Table B. 1: Soil parameters used in exploratory crop simulation models. G iven as 0 - 15, 15 - 30 , 30 - 60, 60 - 90, and 90 - 120cm. Kandeu Lintipe Golomoti Starting water 60%, evenly distrib. 60%, evenly distrib. 60%, evenly distrib. pH 4.95, 4.83, 4.86, 4.89, 4.89 4.88, 4.87, 4.83, 5.13, 5.18 4.7, 4.7, 4.71, 4.7, 4.7 BD (g / cc) 1.53, 1.42, 1.34, 1.4 , 1.4 1.37, 1.5, 1.34, 1.37, 1.28 1.4, 1.4, 1.34, 1.48, 1.65 AirDry (mm / mm) 0.17, 0.18, 0.19, 0.22, 0.22 0.21, 0.23, 0.26, 0.29. 0.32 0.15, 0.18, 0.19, 0.215, 0.25 LL15 (mm / mm) ( low OM soil ) 0.18, 0.24, 0.26, 0.24, 0.24 0.29, 0.32, 0.34, 0.44, 0.43 0 .16, 0.23, 0.29, 0.3, 0.3 LL15 ( high OM soil ) 0.19, 0.24, 0.26, 0.24, 0.24 0.29, 0.32, 0.34, 0.44, 0.43 0.16, 0.23, 0.29, 0.3, 0.3 DUL (mm / mm) ( low OM soil ) 0.28, 0.36, 0.38, 0.37, 0.37 0.41, 0.42, 0.45, 0.48, 0.52 0.28, 0.35, 0.42, 0.42, 0.37 DUL (mm / mm) ( high OM soil ) 0.29, 0.36, 0.38, 0.37, 0.37 0.41, 0.43, 0.45, 0.48, 0.5 0.28, 0.35, 0.42, 0.42, 0.37 SAT (mm / mm) ( low OM soil ) 0.42, 0.46, 0.49, 0.47, 0.47 0.48, 0.43, 0.49, 0.48, 0.52 0.47, 0.47, 0.49, 0.44, 0.38 SAT (mm / mm) ( high OM soil ) 0. 42, 0.46, 0.49, 0.47, 0.47 0.48, 0.43, 0.49, 0.48, 0.52 0.47, 0.47, 0.49, 0.44, 0.38 maize LL (mm / mm) 0.21, 0.23, 0.26, 0.29, 0.32 0.21, 0.23, 0.26, 0.29, 0.32 0.21, 0.23, 0.26, 0.29, 0.32 maize KL 0.08, 0.08, 0.08, 0.08, 0.06 0.08, 0.08, 0.08, 0.08, 0 .06 0.08, 0.08, 0.08, 0.08, 0.06 maize XF 0.7, 0.7, 0.7, 0.4, 0.3 0.5, 0.5, 0.5, 0.2, 0.2 0.4, 0.4, 0.4, 0.3, 0.3 cowpea LL (mm / mm) 0.21, 0.23, 0.26, 0.29, 0.32 0.21, 0.23, 0.26, 0.29, 0.32 0.21, 0.23, 0.26, 0.29, 0.32 cowpea KL 0.08, 0.08, 0.08, 0.08 , 0.06 0.08, 0.08, 0.08, 0.08, 0.06 0.08, 0.08, 0.08, 0.08, 0.06 cowpea XF 0.5, 0.5, 0.5, 0.5, 0.5 0.5, 0.5, 0.5, 0.5, 0.5 0.5, 0.5, 0.5, 0.5, 0.5 Summer Cona: 6.6 6.6 6.6 Summer U: 1.5 1.5 1.5 Summer Date: 1 - Nov 1 - Nov 1 - Nov Winter Cona: 6.6 6.6 6.6 Winter U: 1.5 1.5 1.5 Winter Date: 1 - Apr 1 - Apr 1 - Apr Diffusivity Constant: 88 88 88 Diffusivity Slope: 35.4 35.4 35.4 64 Table B. 1 Kandeu Lintipe Golomoti Soil albedo: 0.2 0.2 0.2 Bare soil runoff curve number: 85 85 85 Max. reduction in c urve number due to cover: 20 20 20 Cover for max curve number reduction: 0.8 0.8 0.8 SWCON 0.2, 0.2, 0.2, 0.2, 0.2 0.2, 0.2, 0.2, 0.2, 0.2 0.2, 0.2, 0.2, 0.2, 0.2 Root C:N ratio 45 45 45 Root weight (kg / ha) 100 100 100 Soil C:N ratio 6.81 14.1 8.06 OC (%) ( low OM soil ) 0.446, 0.312, 0.211, 0.211, 0.211 1.291, 1.028, 1.076, 0.652, 0.395 0.665, 0.653, 0.246, 0.154, 0.208 OC (%) ( high OM soil ) 0.819, 0.76, 0.211, 0.211, 0.211 2.241, 1.894, 1.076, 0.652, 0.395 1.25, 1.228, 0.246, 0.154, 0.208 FBiom 0. 015, 0.015, 0.015, 0.01, 0.01 0.015, 0.015, 0.015, 0.01, 0.01 0.01, 0.01, 0.01, 0.01, 0.01 FInert 0.8, 0.8, 0.8, 0.8, 0.9 0.67, 0.67, 0.67, 0.8, 0.9 0.8, 0.8, 0.9, 0.9, 0.9 Initial NO3 (kg / ha) 1, 1, 1.5, 1.5, 0.25 1, 1, 1.5, 1.5, 0.25 1, 1, 1.5, 1.5, 0 .25 Initial NH4 (kg / ha) 0.1, 0, 0, 0, 0 0.1, 0, 0, 0, 0 0.1, 0, 0, 0, 0 65 Table B. 2: Management parameters used in exploratory crop simulation models. Kandeu Lintipe Golomoti Sow maize using a variable rule with intercropping Name of this crop maiz e maize maize C ultivar : pan6671 r215 pan6671 Method of cropping intercrop intercrop intercrop Exclude from rotation sequence no no no S owing window START date (dd - mmm) 01 - nov 01 - nov 01 - nov S owing window END date (dd - mmm) 30 - dec 30 - dec 30 - dec Must S ow yes yes yes A mount of cumulative rainfall (mm) 15 15 15 N umber of days to accumulate rainfall (days) 3 3 3 A mount of soil water (mm) 15 15 15 S owing opportunity number to sow on (1..n) 1 1 1 U pper limit of sw in top layer (0 - 2) (mm esw/mm soil) 2 2 2 L ower limit of sw in top layer (0 - 1) (mm esw/mm soil) 0 0 0 S owing density (plants/m2) 4 6 4 S owing depth (mm) 30 30 30 Row spacing (m ) 0.75 Amount of starter fertiliser at sowing (kg/ha) ( 69 kg N / ha inter and monocrop ) 23 23 23 Amount of starter fertiliser at sowing (kg/ha) ( 0 kg N / ha inter and monocrop ) 0 0 0 Type urea_N urea_N urea_N N umber of days after sowing to apply top - dress fertiliser (days) 60 72 5 2 Amount of topd ress fertiliser (kg/ha) ( 69 kg N / ha inter and monocrop ) 46 46 46 Amount of t op d ress fertiliser (kg/ha) ( 0 kg N / ha inter and monocrop ) 0 0 0 Type urea_N urea_N urea_N Planter type user_defined user_defined user_defined User_defined depth of seedbed preparation (mm) 150 150 150 F raction of weed population ki lled (0 - 1) 1 1 1 User_defined fraction of surface residues to incorporate (0 - 1) 1 1 1 CN reduction after tillage () : 0 0 0 A mount of cu mulative rain to dissipate tillage effects on cn () : 0 0 0 66 Table B. 2 Kandeu Lintipe Golomoti Sow maize using a variable rule with intercropping N umber of days before window to prepare seed - bed (days) 0 0 0 Must Till (yes/no) no no no Tillage type (user_defined, chisel, disc, planter) disc disc disc User_defined depth of seedbed preparation (mm) 100 100 100 User_defined fraction of surface residues to incorporate (0 - 1) 1 1 1 B iomass removal fraction of growing weeds from field (0 - 1) 0 0 0 CN reduction after tillage () : 0 0 0 A mount of cummulative rain to dissipate tillage effects on CN () : 0 0 0 Sow cowpea using a variable rule with intercropping (only values that differ from the sow maize rule are listed ) Cultivar : spreading spreading spreading S owing window START date (dd - mmm) 1 - jan 1 - jan 1 - jan S owing window END date (dd - mmm) 15 - feb 15 - feb 15 - feb Must Sow yes yes yes A mount of cumulative rainfall (mm) 30 30 30 N umber of days to accumulate rainfall (days) 2 2 2 A mount of soil water (mm) 0 0 0 S owing density (plants/m2) 5 10 5 Amount of starter fertiliser at sowing (kg/ha) ( 69 kg N / h a monocrop ) 23 23 23 Amount of starter fertiliser at sowing (kg/ha) ( 0 kg N / ha monocrop and all intercrops ) 0 0 0 Amount of TopDress fertiliser (kg/ha) ( 69 kg N / ha monocrop ) 46 46 46 Amount of TopDress fertiliser (kg/ha) ( 0 kg N / ha monocrop and all intercrops ) 0 0 0 67 Table B. 2 Kandeu Lintipe Golomoti Harvesting rule for maize F raction of stover to remove (not for maize): 1 1 1 Harvesting rule for cowpea F raction of stover to remove (not for maize): 1 1 1 Tillage with C:N re duction on fixed date T illage date (dd - mmm) : 1 - nov 1 - nov 1 - nov Tillage type : user_defined user_defined user_defined User_defined depth of seedbed preparation (mm) : 15 15 15 User_defined fraction of surface residues to incorporate (0 - 1): .8 .8 .8 CN reduction after tillage () : 10 10 10 A mount of cumulative rain to dissipate tillage effects on CN () 300 300 300 Remove residues at harvest Name of crop harvest to trigger residue removal maize maize maize B iomass removal fraction from field (0 - 1) 1 1 1 Name of crop harvest to trigger residue removal cowpea cowpea cowpea B iomass removal fraction from field (0 - 1) 1 1 1 Reset water, nitrogen and surface OM on fixed date ( only models with soil parameters reset ) D ate of reset (dd - mmm) : 15 - ju l 15 - jul 15 - jul Reset soil water? yes yes yes Reset soil nitrogen? yes yes yes Reset surface organic matter? yes yes yes Table B. 3: Other parameters used in exploratory crop simulation models. Kandeu Lintipe Golomoti Organic matter type maize mai ze maize Initial surface residue 0 0 0 C:N ratio of initial residue 75 75 75 Fraction of residue standing 0 0 0 68 APPENDIX C Outputs from cropping system models run from 1980 2005 in which soil parameters were not re - set annually, allowing for cumulat ive effects. Figure C. 1: Modeled grain yields of cropping systems from 1980 2005 in which soil para meters were not re - set annually, allowing for cumulative effects. Given for the Kandeu (a), Lintipe (b), and Golomoti (c) study location s. Soil org ems abbreviated as Mz mono = maize monocrop, Mz inter = maize intercropped with cowpea, Cp mono = cowpea monocrop, Cp inter = cowpea int ercropped with maize. Bars show mean yield across all years; error bars show one standard error of the mean. 0 2000 4000 6000 8000 Grain yield (kg / ha) a) Mz mono Mz inter Cp mono Cp inter 0 2000 4000 6000 8000 Grain yield (kg / ha) b) 0 2000 4000 6000 8000 Low OM 0N Low OM 69N High OM 0N High OM 69N Grain yield (kg / ha) c) 69 Figure C. 2 : Modeled biomass yields of cropping systems from 1980 2005 in which soil parameters were not re - set annually . Given for the Ka ndeu (a), Lintipe (b), and Golomoti (c) study location s. ems abbreviated as Mz mono = maize monocrop, Mz inter = maize intercro pped with cowpea, Cp mono = cowpea monocrop, Cp inter = cowpea intercropped with maize. Bars show mean yield across all years; error bars show one standard error of the mean. 0 5000 10000 15000 20000 Biomass yield (kg / ha) a) Mz mono Mz inter Cp mono Cp inter 0 5000 10000 15000 20000 Biomass yield (kg / ha) b) 0 5000 10000 15000 20000 Low OM 0N Low OM 69N High OM 0N High OM 69N Biomass yield (kg / ha) c) 70 Figure C. 3 : Probability that average households will meet calorie a nd protein needs with their own production ; contrast of simulations in which soil parameters were re set annually and sim ulations in which parameters were not re set annually . Shown for low organic matter soil at Kandeu with resets (a ) and without resets ( b ), low organic matter soil at Lintipe with resets (c ) and without resets ( d ) , and low organic matter soil at Golomoti with resets (e) and without resets ( f). The chance that a cropping system fails to meet household calorie and protein needs is plotted ag ainst percent of household calorie and protein needs. Therefore, a point at 100% on the y axis and 40% on the x axis indicates that a system fails to meet 100% of household calorie and protein needs 40% of the time 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs a) 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs b) 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs c) 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs d) 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs % chance cal and prot yield < needs e) 0 50 100 150 200 250 300 350 0 25 50 75 100 % of calorie and protein needs % chance cal and prot yield < needs f) 71 WORKS CITED 72 WORKS CITED Adik u, S., Carberry, P., Rose, C., McCown, R., Braddock, R., 1995. A maize (Zea mays) - cowpea (Vigna unguiculata) intercrop model. In: Sinoquet, H., Cruz, P. (Eds.), Ecophysiology of tropical intercropping. INRA Editions, Paris, France, pp. 397 - 406. Adiku, S. , Rose, C., Gabric, A., Braddock, R., Carberry, P., McCown, R., 1998. An evaluation of the performance of maize and cowpea in sole and intercropping systems at two savanna zones of Ghana: A simulation study. Acta Horticulturae 476, 251 - 259. Adiku, S.G.K., Ozier - Lafontaine, H., Bajazet, T., 2001. Patterns of root growth and water uptake of a maize - cowpea mixture grown under greenhouse conditions. Plant and Soil 235, 85 - 94. Altieri, M.A., 2004. Linking ecologists and traditional farmers in the search for sust ainable agriculture. Frontiers in Ecology and the Environment 2, 35 - 42. APSIM Initiative, 2013. APSoil. [ http://www.apsim.info/Products/APSoil.aspx ] Asseng, S., Ewert, F., Rosenzweig, C., Jones, J.W., Hatfield, J.L., Ruane, A.C., Boote, K.J., Thorburn, P.J ., Rotter, R.P., Cammarano, D., Brisson, N., Basso, B., Martre, P., Aggarwal, P.K., Angulo, C., Bertuzzi, P., Biernath, C., Challinor, A.J., Doltra, J., Gayler, S., Goldberg, R., Grant, R., Heng, L., Hooker, J., Hunt, L.A., Ingwersen, J., Izaurralde, R.C., Kersebaum, K.C., Muller, C., Kumar, S.N., Nendel, C., O'Leary, G., Olesen, J.E., Osborne, T.M., Palosuo, T., Priesack, E., Ripoche, D., Semenov, M.A., Shcherbak, I., Steduto, P., Stockle, C., Stratonovitch, P., Streck, T., Supit, I., Tao, F., Travasso, M. , Waha, K., Wallach, D., White, J.W., Williams, J.R., Wolf, J., 2013. Uncertainty in simulating wheat yields under climate change. Nature Climate Change 3, 827 - 832. Bray, R., Kurtz, L., 1945. Determination of total, organic, and available forms of phosphor us in soils. Soil Science 59, 39 - 46. Bryceson, D.F., Fonseca, J., 2006. Risking death for survival: Peasant responses to hunger and HIV/AIDS in Malawi. World Development 34, 1654 - 1666. Byerlee, D., Heisey, P.W., 1996. Past and potential impacts of maize re search in sub - Saharan Africa: A critical assessment. Food Policy 21, 255 - 277. Carberry, P.S., Adiku, S.G.K., McCown, R.L., Keating, B.A., 1996. Application of the APSIM cropping systems model to intercropping systems. In: Ito, O., Johansen, C., Adu - Gyamfi, J.J., Katayama, K., Kumar Rao, J.V.D.K., Rego, T.J. (Eds.), Roots and nitrogen in cropping systems of the semi - arid tropics. International Crops Research Institute for the Semi - Arid Tropics. Carberry, P.S., Muchow, R.C., McCown, R.L., 1989. TESTING THE CE RES - MAIZE SIMULATION - MODEL IN A SEMI - ARID TROPICAL ENVIRONMENT. Field Crops Research 20, 297 - 315. 73 Cash, D.W., Clark, W.C., Alcock, F., Dickson, N.M., Eckley, N., Guston, D.H., Jager, J., Mitchell, R.B., 2003. Knowledge systems for sustainable development. Proceedings of the National Academy of Sciences of the United States of America 100, 8086 - 8091. Chikowo, R., Mapfumo, P., Leffelaar, P.A., Giller, K.E., 2007. Integrating legumes to improve N cycling on smallholder farms in sub - humid Zimbabwe: resource qua lity, biophysical and environmental limitations. In: Bationo, A. (Ed.), Advances in Integrated Soil Fertility Management in Sub - Saharan Africa: Challenges and Opportunities. Springer, London, pp. 231 - 243. Chikowo, R., Mapfumo, P., Nyamugafata, P., Giller, K.E., 2004. Woody legume fallow productivity, biological N - 2 - fixation and residual benefits to two successive maize crops in Zimbabwe. Plant and Soil 262, 303 - 315. Chinsinga, B., 2004. Poverty and food security in Malawi: Some policy reflections on the con text of crumbling traditional support systems. Canadian Journal of Development Studies - Revue Canadienne D Etudes Du Developpement 25, 321 - 341. Chirwa, E., Kydd, J., Dorward, A., 2006. Future scenarios for agriculture in Malawi: challenges and dilemmas. In: D, H. (Ed.), Future Agricultures. Future Agricultures Consortium. Connolly, J., 1986. ON DIFFICULTIES WITH REPLACEMENT - SERIES METHODOLOGY IN MIXTURE EXPERIMENTS. Journal of Applied Ecology 23, 125 - 137. Denning, G., Kabambe, P., Sanchez, P., Malik, A., Flo r, R., Harawa, R., Nkhoma, P., Zamba, C., Banda, C., Magombo, C., Keating, M., Wangila, J., Sachs, J., 2009. Input Subsidies to Improve Smallholder Maize Productivity in Malawi: Toward an African Green Revolution. Plos Biology 7, 2 - 10. Drost, Unpublished d ata. Elowad, H.O.A., Hall, A.E., 1987. INFLUENCES OF EARLY AND LATE NITROGEN - FERTILIZATION ON YIELD AND NITROGEN - FIXATION OF COWPEA UNDER WELL - WATERED AND DRY FIELD CONDITIONS. Field Crops Research 15, 229 - 244. Faris, M. A., Burity, H. A., Dos Reis, O. V., & Mafra, R. C., 1983. Intercropping of sorghum or maize with cowpeas or common beans under two fertility regimes in Northeastern Brazil. Experimental Agriculture 19 , 251 - 261. Food and Agriculture Organization, 2004. Human energy requirements: Report of a Joint FAO/WHO/UNU Expert Consultation., FAO Food and Nutrition Technical Report Series, Rome, Italy. Fukai, S., Trenbath, B.R., 1993. PROCESSES DETERMINING INTERCROP PRODUCTIVITY AND YIELDS OF COMPONENT CROPS. Field Crops Research 34, 247 - 271. 74 Ghosh, P.K., Mohanty, M., Bandyopadhyay, K.K., Painuli, D.K., Misra, A.K., 2006. Growth, competition, yield advantage and economics in soybean/pigeonpea intercropping system in semi - arid tropics of India - I. Effect of subsoiling. Field Crops Research 96, 80 - 89. Gille r, K.E., Rowe, E.C., de Ridder, N., van Keulen, H., 2006. Resource use dynamics and interactions in the tropics: Scaling up in space and time. Agricultural Systems 88, 8 - 27. Giller, K.E., Witter, E., Corbeels, M., Tittonell, P., 2009. Conservation agricult ure and smallholder farming in Africa: The heretics' view. Field Crops Research 114, 23 - 34. climate change: Multinomial choice analysis. African Journal of Agricu ltural and Resource Economics 2.1, 83 - 104. Hendershot, W., LaLande, H., Douquette, M., 1993. Soil pH in water. In: Carter, M. (Ed.), Soil Sampling and Methods of Analysis. CRC Press, Boca Raton, FL. Hildebrand, P., Russell, J., 1996. Adaptability analysis: A method for the design, analysis and interpretation of on - farm research - extension. Iowa State University Press, Ames, Iowa. Hooper, D.U., Chapin, F.S., Ewel, J.J., Hector, A., Inchausti, P., Lavorel, S., Lawton, J.H., Lodge, D.M., Loreau, M., Naeem, S., Schmid, B., Setala, H., Symstad, A.J., Vandermeer, J., Wardle, D.A., 2005. Effects of biodiversity on ecosystem functioning: A consensus of current knowledge. Ecological Monographs 75, 3 - 35. Hulugalle, N.R., Lal, R., 1986. SOIL - WATER BALANCE OF INTERCROPPE D MAIZE AND COWPEA GROWN IN A TROPICAL HYDROMORPHIC SOIL IN WESTERN NIGERIA. Agronomy Journal 78, 86 - 90. Jackai, L., Adalla, C., 1997. Pest management practices in cowpea: a review. In: Singh, B., Mohan Raj, D., Dashiell, K., Jackai, L. (Eds.), Advances in Cowpea Research. International Institute for Tropical Agriculture, Ibadan, Nigeria, pp. 240 - 258. Jones, A., Shrinivas, R., Bezner - Kerr, R., 2014. Farm production diversity is associated with greater household dietary diversity in Malawi: Findings from nat ionally representative data. Food Policy 46, 1 - 12. Kamanga, B.C.G., Waddington, S.R., Robertson, M.J., Giller, K.E., 2010. RISK ANALYSIS OF MAIZE - LEGUME CROP COMBINATIONS WITH SMALLHOLDER FARMERS VARYING IN RESOURCE ENDOWMENT IN CENTRAL MALAWI. Experimenta l Agriculture 46, 1 - 21. Keating, B.A., Carberry, P.S., 1993. RESOURCE CAPTURE AND USE IN INTERCROPPING - SOLAR - RADIATION. Field Crops Research 34, 273 - 301. Keating, B.A., Carberry, P.S., Hammer, G.L., Probert, M.E., Robertson, M.J., Holzworth, D., Huth, N. I., Hargreaves, J.N.G., Meinke, H., Hochman, Z., McLean, G., Verburg, K., Snow, V., Dimes, J.P., Silburn, M., Wang, E., Brown, S., Bristow, K.L., Asseng, S., Chapman, S., 75 McCown, R.L., Freebairn, D.M., Smith, C.J., 2003. An overview of APSIM, a model desig ned for farming systems simulation. European Journal of Agronomy 18, 267 - 288. Kellogg Biological Station LTER, 2003. Carlo Erba NA1500 Series II Combustion Analyzer. Kellogg Biological Station LTER, 2008. KBS058: Particle Size Analysis for Soil Texture Det ermination (hydrometer method). Kerr, J., Kolavalli, S., 1999. Impact of agricultural research on poverty alleviation: conceptual framework with illustrations from the literature. EPTD Discussion Paper. IFPRI. Kurtz, T., Melsted, H., Bray, R., 1952. The im portance of nitrogen and water in reducing competition between intercrops and corn. Agronomy Journal 44, 13 - 17. Lightfoot, C.W.F., Dear, K.B.G., Mead, R., 1987. INTERCROPPING SORGHUM WITH COWPEA IN DRYLAND FARMING SYSTEMS IN BOTSWANA .2. COMPARATIVE STABIL ITY OF ALTERNATIVE CROPPING SYSTEMS. Experimental Agriculture 23, 435 - 442. Lobell, D.B., Burke, M.B., Tebaldi, C., Mastrandrea, M.D., Falcon, W.P., Naylor, R.L., 2008. Prioritizing climate change adaptation needs for food security in 2030. Science 319, 607 - 610. Lowole, M., 1984. Soil Map of Malawi. Malawi Department of Agricultural Research, Lilongwe, Malawi. Mabapa, P.M., Ogola, J.B.O., Odhiambo, J.J.O., Whitbread, A., Hargreaves, J., 2010. Effect of phosphorus fertilizer rates on growth and yield of three soybean (Glycine max) cultivars in Limpopo Province. African Journal of Agricultural Research 5, 2653 - 2660. Malezieux, E., Crozat, Y., Dupraz, C., Laurans, M., Makowski, D., Ozier - Lafontaine, H., Rapidel, B., de Tourdonnet, S., Valantin - Morison, M., 2009. Mixing plant species in cropping systems: concepts, tools and models. A review. Agronomy for Sustainable Development 29, 43 - 62. McDougall, C., Braun, A., 2003. Navigating complexity, diversity and dynamism: reflections on research for natural resource man agement. In: Pound, B., Snapp, S., McDougall, C., Braun, A. (Eds.), Managing natural resources for sustainable livelihoods: Uniting science and participation. Earthscan, UK/USA and International Development Research Centre (IDRC), Ottawa, pp. 20 - 44. Mhango , W.G., Snapp, S.S., Phiri, G.K., 2012. Opportunities and constraints to legume for sustainable maize production on smallholder farms in Malawi. Renewable Agriculture and Food Systems. Morris, R.A., Garrity, D.P., 1993. RESOURCE CAPTURE AND UTILIZATION IN INTERCROPPING - WATER. Field Crops Research 34. 76 Mortimore, M., Singh, B., Harris, F., Blade, S., 1997. Cowpea in traditional cropping systems. In: Singh, B., Mohan Raj, D., Dashiell, K., Jackai, L. (Eds.), Advances in Cowpea Research. Intern ational Institute for Tropical Agriculture, Ibadan, Nigeria. Murage, E.W., Karanja, N.K., Smithson, P.C., Woomer, P.L., 2000. Diagnostic indicators of soil quality in productive and non - Agricul ture, Ecosystems & Environment 79, 1 - 8. Nelson, R.A., Holzworth, D.P., Hammer, G.L., Hayman, P.T., 2002. Infusing the use of seasonal climate forecasting into crop management practice in North East Australia using discussion support software. Agricultural Systems 74, 393 - 414. Ngwira, L., Mkandawire, A., Edje, O., Tinsley, R., 1990. Malawi Experiences in Intercropping Research. In: Waddington, S., Palmer, A., Edje, O. (Eds.), Research Methods for Cereal/legume Intercropping: Proceedings of a Workshop on Res earch Methods for Cereal/Legume Intercropping in Eastern and Southern Africa Held at Lilongwe, Malawi, 23 - 27 January 1989 (No. 17). CIMMYT. Nielsen, S., Ohler, T., Mitchell, C., 1997. Cowpea leaves for human consumption: production, utilization and nutrie nt composition. In: Singh, B., Mohan Raj, D., Dashiell, K., Jackai, L. (Eds.), Advances in Cowpea Research. International Institute for Tropical Agriculture, Ibadan, Nigeria, pp. 326 - 332. OECD/FAO, 2012. OECD - FAO Agricultural Outlook 2012 - 2021. OECD Publis hing and FAO. Ofori, F., Stern, W.R., 1987. THE COMBINED EFFECTS OF NITROGEN - FERTILIZER AND DENSITY OF THE LEGUME COMPONENT ON PRODUCTION EFFICIENCY IN A MAIZE COWPEA INTERCROP SYSTEM. Field Crops Research 16, 43 - 52. Ogindo, H.O., Walker, S., 2005. Compari son of measured changes in seasonal soil water content by rainfed maize - bean intercrop and component cropping systems in a semi - arid region of southern Africa. Physics and Chemistry of the Earth 30, 799 - 808. Ollenberger, M., 2012. Monitoring integrated soi l fertility management options in Malawi under variable soil and climate conditions. Masters thesis, Crop and Soil Sciences. Michigan State University, East Lansing, MI. Ozor, N., Cynthia, N., 2010. Difficulties in adaptation to climate change by farmers i n Enugu state, Nigeria. Journal of Agricultural Extension 14, 106 - 122. Padulosi, S., Ng, N., 1997. Origin, taxonomy, and morphology of Vigna unguiculata (L.) Walp. In: Singh, R., Mohan Raj, D., Dashiell, K., Jackai, L. (Eds.), Advances in Cowpea Research. International Institute for Tropical Agriculture, Ibadan, Nigeria, pp. 1 - 12. Peters, P.E., 2002. The limits of knowledge: securing rural livelihoods in a situation of resource scarcity. In: Barrett, C.B., Place, F., Aboud, A.A. (Eds.), Natural resources ma nagement in African agriculture: understanding and improving current practices. CABI Publishing, New York, pp. 35 - 50. 77 Probert, M.E., Dimes, J.P., Keating, B.A., Dalal, R.C., Strong, W.M., 1998. APSIM's water and nitrogen modules and simulation of the dynam ics of water and nitrogen in fallow systems. Agricultural Systems 56, 1 - 28. Robertson, M.J., Sakala, W., Benson, T., Shamudzarira, Z., 2005. Simulating response of maize to previous velvet bean (Mucuna pruriens) crop and nitrogen fertiliser in Malawi. Fiel d Crops Research 91, 91 - 105. Rusinamhodzi, L., Corbeels, M., Nyamangara, J., Giller, K., 2012. Maize - grain legume intercropping is an attractive option for ecological intensification that reduces climate risk for smallholder farmers in central Mozambique. Field Crops Research 136, 12 - 22. Saha, S., Moorthi, S., Pan, H.L., Wu, X.R., Wang, J.D., Nadiga, S., Tripp, P., Kistler, R., Woollen, J., Behringer, D., Liu, H.X., Stokes, D., Grumbine, R., Gayno, G., Wang, J., Hou, Y.T., Chuang, H.Y., Juang, H.M.H., Sela, J., Iredell, M., Treadon, R., Kleist, D., Van Delst, P., Keyser, D., Derber, J., Ek, M., Meng, J., Wei, H.L., Yang, R.Q., Lord, S., Van den Dool, H., Kumar, A., Wang, W.Q., Long, C., Chelliah, M., Xue, Y., Huang, B.Y., Schemm, J.K., Ebisuzaki, W., Lin, R. , Xie, P.P., Chen, M.Y., Zhou, S.T., Higgins, W., Zou, C.Z., Liu, Q.H., Chen, Y., Han, Y., Cucurull, L., Reynolds, R.W., Rutledge, G., Goldberg, M., 2010. THE NCEP CLIMATE FORECAST SYSTEM REANALYSIS. Bulletin of the American Meteorological Society 91, 1015 - 1057. Saha, S., Moorthi, S., Wu, X., Wang, J., Nadiga, S., Tripp, P., Behringer, D., Hou, Y., Chuang, H., Iredell, M., Ek, M., Meng, J., Yang, R., Mendez, M., van den Dool, H., Zhang, Q., Wang, W., Chen, M., Becker, E., 2014. The NCEP Climate Forecast Sys tem Version 2. Journal of Climate 27, 2185 - 2208. Saxton, K.E., Rawls, W.J., 2006. Soil water characteristic estimates by texture and organic matter for hydrologic solutions. Soil Science Society of America Journal 70, 1569 - 1578. Scholes, R.J., Biggs, R., 2 004. Ecosystem services in southern Africa: a regional assessment. Millennium Ecosystem Assessment. Council for Scientific and Industrial Research, Pretoria, South Africa. Searle, P.G.E., Comudom, Y., Shedden, D.C., Nance, R.A., 1981. EFFECT OF MAIZE + LEG UME INTERCROPPING SYSTEMS AND FERTILIZER NITROGEN ON CROP YIELDS AND RESIDUAL NITROGEN. Field Crops Research 4, 133 - 145. Shaxson, L., Tauer, L.W., 1992. INTERCROPPING AND DIVERSITY - AN ECONOMIC - ANALYSIS OF CROPPING PATTERNS ON SMALLHOLDER FARMS IN MALAWI. Experimental Agriculture 28, 211 - 228. Shumba, E., Dhliwayo, H., Mukoko, O., 1990. The potential of maize - cowpea intercropping in low rainfall areas of Zimbabwe. Zimbabwe Journal of Agricultural Research 28, 33 - 38. Sileshi, G.W., Akinnifesi, F.K., Ajayi, O .C., Muys, B., 2011. Integration of legume trees in maize - based cropping systems improves rain use efficiency and yield stability under rain - fed agriculture. Agricultural Water Management 98. 78 Sinclair, T.R., Seligman, N., 2000. Criteria for publishing pape rs on crop modeling. Field Crops Research 68, 165 - 172. Skovgard, H., Pats, P., 1996. Effects of intercropping on maize stem borers and their natural enemies. Bulletin of Entomological Research 86, 599 - 607. Sleutel, S., De Neve, S., Singier, B., Hofman, G., 2007. Quantification of organic carbon in soils: A comparison of methodologies and assessment of the carbon content of organic matter. Communications in Soil Science and Plant Analysis 38, 2647 - 2657. Snapp, S., Bezner Kerr, R., Smith, A., Ollenberger, M., Mhango, W., Shumba, L., Gondwe, T., Kanyama - Phiri, G., 2014. Modeling and participatory, farmer - led approaches to food security in a changing world: a case study from Malawi. Secheresse in press. Snapp, S., Kanyama - Phiri, G., Kamanga, B., Gilbert, R., Wel lard, K., 2002a. Farmer and researcher partnerships in Malawi: Developing soil fertility technologies for the near - term and far - term. Experimental Agriculture 38, 411 - 431. Snapp, S.S., 1998. Soil nutrient status of smallholder farms in Malawi. Communicatio ns in Soil Science and Plant Analysis 29, 2571 - 2588. Snapp, S.S., Blackie, M.J., Gilbert, R.A., Bezner - Kerr, R., Kanyama - Phiri, G.Y., 2010. Biodiversity can support a greener revolution in Africa. Proceedings of the National Academy of Sciences of the Unit ed States of America 107, 20840 - 20845. Snapp, S.S., Mafongoya, P.L., Waddington, S., 1998. Organic matter technologies for integrated nutrient management in smallholder cropping systems of southern Africa. Agriculture Ecosystems & Environment 71, 185 - 200. Snapp, S.S., Rohrbach, D.D., Simtowe, F., Freeman, H.A., 2002b. Sustainable soil management options for Malawi: can smallholder farmers grow more legumes? Agriculture Ecosystems & Environment 91, 159 - 174. Thierfelder, C., Cheesman, S., Rusinamhodzi, L., 20 12. A comparative analysis of conservation agriculture systems: Benefits and challenges of rotations and intercropping in Zimbabwe. Field Crops Research 137, 237 - 250. Thomalla, F., Downing, T., Spanger - Siegfried, E., Han, G.Y., Rockstrom, J., 2006. Reducin g hazard vulnerability: towards a common approach between disaster risk reduction and climate adaptation. Disasters 30, 39 - 48. Thomas, D.S.G., Twyman, C., 2005. Equity and justice in climate change adaptation amongst natural - resource - dependent societies. G lobal Environmental Change - Human and Policy Dimensions 15, 115 - 124. Thornton, P.K., Jones, P.G., Ericksen, P.J., Challinor, A.J., 2011. Agriculture and food systems in sub - Saharan Africa in a 4 degrees C+ world. Philosophical Transactions of the Royal Soci ety a - Mathematical Physical and Engineering Sciences 369, 117 - 136. 79 Toenniessen, G., Adesina, A., DeVries, J., 2008. Building an Alliance for a Green Revolution in Africa. Reducing the Impact of Poverty on Health and Human Development: Scientific Approaches 1136, 233 - 242. Trenbath, B.R., 1999. Multispecies cropping systems in India - Predictions of their productivity, stability, resilience and ecological sustainability. Agroforestry Systems 45, 81 - 107. Twomlow, S., Mugabe, F.T., Mwale, M., Delve, R., Nanja, D., Carberry, P., Howden, M., 2008. Building adaptive capacity to cope with increasing vulnerability due to climatic change in Africa - A new approach. Physics and Chemistry of the Earth 33, 780 - 787. van den Belt, M., 2004. Mediated Modeling: A system dyna mics approach to environmental consensus building. Island Press, Washington DC. Vandermeer, J., 1995. THE ECOLOGICAL BASIS OF ALTERNATIVE AGRICULTURE. Annual Review of Ecology and Systematics 26, 201 - 224. Vanlauwe, B., Bationo, A., Chianu, J., Giller, K.E. , Merckx, R., Mokwunye, U., Ohiokpehai, O., Pypers, P., Tabo, R., Shepherd, K.D., Smaling, E.M.A., Woomer, P.L., Sanginga, N., 2010. Integrated soil fertility management Operational definition and consequences for implementation and dissemination. Outlook on Agriculture 39, 17 - 24. Vesterager, J., Nielsen, N., Hogh - Jensen, H., 2008. Effects of cropping history and phosphorus source on yield and nitrogen fixation in sole and intercropped cowpea - maize systems. Nutrient Cycling in Agroecosystems 80, 61 - 73. Wadd ington, S.R., Mekuria, M., Siziba, S., Karigwindi, J., 2007. Long - term yield sustainability and financial returns from grain legume - maize intercrops on a sandy soil in subhumid north central Zimbabwe. Experimental Agriculture 43, 489 - 503. Watiki, J.M., Fuk ai, S., Banda, J.A., Keating, B.A., 1993. RADIATION INTERCEPTION AND GROWTH OF MAIZE/COWPEA INTERCROP AS AFFECTED BY MAIZE PLANT - DENSITY AND COWPEA CULTIVAR. Field Crops Research 35, 123 - 133. Whitbread, A.M., Robertson, M.J., Carberry, P.S., Dimes, J.P., 2 010. How farming systems simulation can aid the development of more sustainable smallholder farming systems in southern Africa. European Journal of Agronomy 32, 51 - 58. Wood, A., Mayer, J., 2006. Malawi poverty and vulnerability assessment: Investing in our future. The Republic of Malawi and The World Bank. World Health Organization, 2007. Protein and amino acid requirements in human nutrition: Report of a Joint WHO/FAO/U NU expert consultation. WHO Technical Report Series. World Health Organization, Geneva, Switzerland. Xiong, W., Holman, I., Conway, D., Lin, E., Li, Y., 2008. A crop model cross calibration for use in regional climate impacts studies. Ecological Modelling 213, 365 - 380. 80 Yamoah, C.F., Walters, D.T., Shapiro, C.A., Francis, C.A., Hayes, M.J., 2000. Standardized precipitation index and nitrogen rate effects on crop yields and risk distribution in maize. Agriculture Ecosystems & Environment 80, 113 - 120.