diff --git "a/main/part_2/0132734289.json" "b/main/part_2/0132734289.json"
new file mode 100644--- /dev/null
+++ "b/main/part_2/0132734289.json"
@@ -0,0 +1 @@
+{"metadata":{"gardian_id":"27b8ca19ee5e8aae92c69195d9644c46","source":"gardian_index","url":"https://cgspace.cgiar.org/rest/bitstreams/bc972c59-ceb4-4d8a-aa20-714ffbf5bff1/retrieve","id":"-1620388450"},"keywords":[],"sieverID":"d9ba1f78-baba-4ef4-9ae3-0b1a2e0f9c51","content":"Maize as food, feed and fertiliser in intensifying crop-livestock systems in East and southern Africa: An ex ante impact assessment of technology interventions to improve smallholder welfare.varieties as a percentage of total area planted to maize Table 2. Average nutrient balance of NPK (kg/ha per year) of arable land in selected countries Table 3. Multiple cropping combinations including maize found in four different areas of Malawi Table 4. Maize cultivation and manure use on farms in selected districts of some countries of East and southern Africa Table 5. Responses in yields of maize grain (t/ha) to a mixed application of manure and fertiliser in Zambia Table 6. Maize statistics for the countries of East and southern Africa (1980-96) Table 7. Summary of the key features of the case study locations in Kenya Table 8. Summary of the key features of the case study locations in Malawi Table 9. Summary of the key features of the case study locations in the Republic of South Africa Table 10. Summary of the key features of the case study locations in Tanzania Table 11. Summary of the key features of the case study locations in Zimbabwe Table 12. The potential outcomes of a range of technological interventions for maize-based farming systems Table 13. Defining variables used to assess the impacts of technologies in different production outcome groups Table 14. Summary classification of maize-livestock systems based on descriptors derived from the country case studies Table 15. Relationships between maize-livestock and Food and Agriculture Organisation (FAO) system classifications Table 16. Total human populations (2000) associated with each population density class for the five case study countries Table 17. Distribution of population by the nature of the cropping season Table 18. Technology-system matching matrix for small-scale, intensive systems Table 19. Technology-system matching matrix for medium-scale, intensive systems Table 20. Technology-system matching matrix for medium-scale, semiintensive systems Table 21. Technology-system matching matrix for medium-scale, extensive systems Table 22. Potential interventions relating to the maize crop for food and feed use and their likely areas of impact Table 23. Baseline data for the small-scale, intensive production system Table 24. Baseline data for the medium-scale, intensive production system Table 25. Baseline data for the medium-scale, semi-intensive production system Table 26. Baseline data for the medium-scale, extensive production system Table 27. Predicted impacts of weed collection from the maize crop in small-scale, intensive farming systems Table 28. Predicted impacts of weed collection from the maize crop in medium-scale, intensive farming systems Table 29. Predicted impacts of improved management of green maize stover in small-scale, intensive farming systems Table 30. Predicted impacts of improved management of green maize stover in medium-scale, intensive farming systems Table 31. Predicted impacts of improved feeding systems incorporating dry maize stover in small-scale, intensive farming systems Table 32. Predicted impacts of improved feeding systems incorporating dry maize stover in medium-scale, intensive farming systems Table 33. Predicted impacts of improved feeding systems incorporating dry maize stover in medium-scale, semi-intensive farming systems Table 34. Predicted impacts of improved feeding systems incorporating dry maize stover in medium-scale, extensive farming systems Table 35. Predicted impacts of chopping dry maize stover in medium-scale, intensive farming systems Table 36. Predicted impacts of adopting replacement fodder crops in small-scale, intensive farming systems Table 37. Predicted impacts of adopting intercropping in medium-scale, semi-intensive farming systems Table 38. Predicted impacts of adopting intercropping in medium-scale, extensive farming systems Table 39. Predicted impacts of improved manure management strategies in small-scale, intensive farming systems Table 40. Predicted impacts of improved manure management strategies in medium-scale, intensive farming systems Table 41. Impacts of a 5% improvement in maize stover digestibility on individual milk production from dairy cattle in the small-and medium-scale, intensive production systems of the five countries Table 42. Human population, cattle population, cultivated area and estimated maize area by country by system Table 43. Location of tables describing baseline data, predicted impacts and aggregated impacts, by system Table 44. Use of collected weeds of the maize crop for livestock feeding in small-scale, intensive systems Table 45. Improved management of green maize stover for feed use in small-scale, intensive systems Table 46. Improved feeding systems incorporating dry maize stover in small-scale, intensive systems Table 47. Adoption of replacement fodder crops in small-scale, intensive systems Table 48. Improved manure management strategies in small-scale, intensive systems Table 49. Use of collected weeds of the maize crop for livestock feeding in medium-scale, intensive systems v Table 50. Improved management of green maize stover for feed use in medium-scale, intensive systems Table 51. Improved feeding systems incorporating dry maize stover in medium-scale, intensive systems Table 52. Chopping / soaking of dry maize stover in medium-scale, intensive systems Table 53. Improved manure management strategies in medium-scale, intensive systems Table 54. Improved feeding systems incorporating dry maize stover in medium-scale, semi-intensive systems Table 55. Adoption of intercropping in medium-scale, semi-intensive systems Table 56. Improved feeding systems incorporating dry maize stover in medium-scale, extensive systems Table 57. Adoption of intercropping in medium-scale, extensive systems Table 58. Potential impacts of improved stover quality in the small-and medium-scale systems of the five countries Table 59. Value of the defining variables used to assess the impacts of technologies in different production outcome groups Table 60. Potential interventions: research and extensions costs and adoption parameters Table 61. Parameters for the demand and supply curves for maize, meat and milk, and country production figures for these commodities for 2000 Table 62. Incremental costs of production by intervention and system Table 63. Baseline analysis over 20 years for the interventions by system Table 64. Selected sensitivity analysis: present value of net benefit streams to 2020 (US$ million) Table 65. Selected sensitivity analysis: interventions by system ranked by present value of the net benefit stream to 2020 Table 66. Comparison of the study countries and all the countries of East and southern Africa in terms of the mixed maize-based systems Table 67. Summary of feasible interventions by system, with an indication of the research and/or extension effort needed to realise the potential net benefits viFigure 1. Location of study sites and associated town, district or province centroids from the country studies Figure 2. An outline of the relationships between system components and intervention groups in maize-based farming systems Figure 3. Farming systems in mixed maize areas in five countries of East and southern Africa Figure 4. Cattle density in mixed maize areas in five countries of East and southern Africa Figure 5. Human population density in mixed maize areas in five countries of East and southern Africa, 2000 Figure 6. Human population density in mixed maize areas in five countries of East and southern Africa, 2020 Figure 7. Parameters required for each research intervention Figure 8. Farming systems in mixed maize areas in all the countries of East and southern Africa Figure 9. Human population density in mixed maize areas in all the countries of East and southern Africa, 2000 Figure 10. Human population density in mixed maize areas in all the countries of East and southern Africa, 2020viii Executive SummaryThis report describes an ex ante impact assessment study carried out between 1999 and 2001 to generate information to help guide future research activities in East and southern Africa concerning the maize crop in the mixed systems of the region. Maize is the staple food for perhaps 25 million households in the region; it is planted annually on more than 15 million hectares of land and contributes at least 25% of the calories to the diets of more than 80 million people. Much of this maize is grown in the mixed systems of the region, which contain 170 million people. This number is expected to grow to 266 million people by 2020, so the issues of food security and sustainable livelihoods for this burgeoning population are of great importance and concern.Country studies were conducted in Kenya, Tanzania, Malawi, Zimbabwe and the Republic of South Africa (RSA) to characterise the maize-based, mixed systems in the region. From this work, four systems were identified: small-scale intensive (SSI), medium-scale intensive (MSI), medium-scale semi-intensive (MSSI) and medium-scale extensive (MSE). These were defined in terms of human population density and maize cropping density, and subsequently mapped for the five countries. From the many crop and livestock management options that could be explored, information from the country case studies and the literature was used to define a subset of alternatives that either are in use in some of the systems and countries of the region but not in all, or show particular promise, given smallholders' constraints and attitudes, for increasing incomes and productivity. These were:• use of collected weeds of the maize crop for livestock feeding programmes • improved management of green maize stover for feed use • improved feeding systems incorporating dry maize stover • chopping/soaking of dry maize stover • use of replacement fodder crops • intercropping • improved manure management strategies and • selection and/or breeding for improved digestibility of maize stover.The productivity impacts of these system changes were assessed using crop and livestock simulation models. The household-level impacts by system within the various countries in which the interventions were judged to be relevant were then aggregated, using the geographic information systems (GIS) data layers derived in the characterisation stage. Each intervention-by-system combination was then assessed, using the economic surplus model, in relation to the potential impacts that could arise as a result of resource expenditures on research and extension to develop and disseminate the particular intervention. The results indicated that improved feeding systems offer substantial potential for smallholders, particularly in the more intensive systems. Promoting the use of intercropping in the more extensive systems where this is not already practised also offers substantial net benefits. The use of weeds, and improved green stover management, are interventions with modest research and extension costs but these provide significant net benefits in the more intensive systems. There may also be significant returns to improved manure management but these are particularly sensitive to the assumptions made in the analysis. The analysis identified viable options for all system types. Three features were apparent: the number of viable options decreases as system intensity decreases; the amount of research effort that is needed decreases as system intensity decreases; and the amount of extension effort needed for many of the options is consistently high, whichever systems are being targeted. This highlights the importance of extension, and its general treatment as a 'poor relation' of 'proper' research is probably a major constraint to the widespread adoption of perfectly feasible technologies.Much could be done to improve the analysis, particularly the longer-term impacts of some of the interventions. However, despite their limitations, we now have the tools to start to assess different interventions at the system level. The analysis has provided insights into the nature of the interventions that could assist in improving the lot of smallholders in the region who depend for their livelihood on maize-based, crop-livestock systems now and in the future.With increasing human population density and improved market access for the sale of the crop and livestock products, smallholder agricultural systems in East and southern Africa are responding by intensifying production. The intensification of livestock production may involve the adoption of a dairy enterprise, which can contribute significantly to the income of producer households and the welfare of rural communities. This is also an effective means by which plant nutrients can be rapidly recycled within and between farms. The cattle herd can also serve as a source of draft animals. Both these roles can improve the management of natural resources on smallholder farms. The factors driving intensification, however, often lead to the expansion of cropped areas and more intensive cropping practices at the expense of grazing land. In the face of declining grazing land, the potential of arable land to provide fodder throughout the year must be enhanced, if the important role of livestock within the farm system and for household welfare is to be maintained or developed.Maize is the staple food of 24 million households in East and southern Africa and is annually planted over an area of 15.5 million hectares. Research into agronomic practices to optimise grain yields is a priority for governments in the region because of the critical role played by maize in food security. As a result, agronomic evaluation and crop husbandry recommendations for maize focus on optimising plant population density and reducing weed competition for maximising grain yield, but have generally paid little attention to the maize crop (or weeds) as sources of cattle fodder.Research on the dry residues of maize (stover) has centred on how best to make use of what is essentially a low quality, seasonally available fodder and on strategies to better exploit the use of residues in soil fertility maintenance. However, research has failed to adequately explain the lack of adoption of more comprehensive maize residue use for livestock feed or soil amendment on smallholder farms, and to analyse the economic and environmental consequences of these competing uses for maize residues. As maize-based, crop-livestock systems intensify, the challenge is to ensure that maize production contributes more food for humans and more fodder for ruminant livestock, thereby improving food security while at the same time protecting the natural resource base.Various research activities in the region are currently concentrating on the participatory development of management practices for maize that contribute to more productive crop-livestock (particularly dairy) farming through higher and more stable biomass (grain and fodder) yields, while protecting the natural resource base. As part of these efforts, the System-wide Livestock Programme (SLP) of the Consultative Group on International Agricultural Research (CGIAR), convened by ILRI, supported an ex ante impact assessment of the potential role of maize residues in the region. This gives particular emphasis on extending the recommendation domain for the results through learning from indigenous husbandry practices at sites selected to reflect contrasting human population densities, market access and climate within maize-based, intensified crop-cattle systems. The objective of this impact assessment was to be able to address the constraints to increased use of maize as fodder within the context of farmer experience and knowledge within a continuum of intensification of maizelivestock systems.The original project proposal was formulated in various modules, to address:• Identification of the biological and socio-economic factors necessary for adoption of alternative practices for maize residue use at farm level • Husbandry strategies for sustainable increases in food and feed biomass production from maize • Enhanced simulation models to assess impact of modified maize husbandry practices on yield of food/feed and soil fertility and • Measured impact of modified cropping practices on long-term soil fertility.The original research activities were arranged into four components, described briefly below.This activity was made up of (1) appraisals of farmers' preferences and studies of adoption of practices related to the use of biomass from maize and companion crops on farms at each of the six sites; and (2) an ex ante economic impact assessment of viable research interventions for five countries in the region. An initial workshop was held in Harare in late 1998 and work on the ex ante impact assessment began in earnest in 1999. It was envisaged that the study proper would commence once the ex ante work had been completed, involving benchmark sites characterised by human population density, market access, climatic factors and farm structure. In this way, it was anticipated that the research outputs would be representative of defined recommendation domains, ensuring that the outputs can be extrapolated to other regions with comparable, or anticipated levels of, intensification of smallholder agriculture.The work proceeded in various phases. First, a consultant was commissioned to carry out a literature review of maize in smallholder systems in the region (Reynolds 1999a). Second, with input from the Harare workshop of 1998, a set of country studies was commissioned from three consultants. Countries covered were RSA (Le Roux 1999), Zimbabwe and Malawi (Reynolds 1999b), Tanzania (Thorne 1999a) and Kenya (Thorne 1999b). Although the analysis described here could easily be extended to other countries in the region (such as Ethiopia and Uganda, for example), initial SLP resources were available to cover only these five. These reports generated a great deal of information on the potential value of the returns to planned research aimed at improving cropping methods for greater food and feed production in the smallholder maize-livestock systems in the target countries. An important part of this activity is the identification of recommendation domains through overlaying and analysis of existing spatial maize and livestock distribution databases with other biophysical and socio-economic indicators, allowing the extrapolation of results and the better targeting of management practices both within the region and trans-regionally.The objective of this component was to construct frameworks based on resource flow diagrams to allow crop-livestock interactions to be assessed. These would incorporate agronomic, animal nutrition and socio-economic components of maize production and use at selected pilot study sites. These resource and nutrient flows were then quantified using information from the literature and, where appropriate, existing crop and livestock simulation models or simple, spreadsheet-based, planning tools developed for the project. The resource frameworks would then be used to elucidate key interactions in the maizelivestock-soil continuum: nutrient flows, complementarities and trade-offs inherent in alternative management strategies, and the sensitivity of producers to risk. Further testing of the approaches developed was to be carried out at a suitable site or sites identified by activities under Component 1. The tools developed through activities carried out under this component could then be used for ex ante impact assessment relating to interventions studied in later components of the project.To date, these activities have been carried out at Embu, in concert with other Kenya Agricultural Research Institute (KARI) and ILRI activities there. The Embu site has been used as the initial test bed for the development of the required tools as data are currently being collected under other projects. These prototype tools will in the future be tested and refined at other sites in the region.1.3 Component 3. Management, genetic and environmental influences on production and feeding value of maize fodderThis component was designed to evaluate the management, genetic and environmental influences on maize grain and fodder yield and quality. It was expected that the results of the ex ante impact assessment would indicate which maize-specific strategies needed to be investigated in on-farm experiments conducted at several locations over consecutive cropping seasons. For example, local, improved and promising maize varieties would need to be evaluated for grain yield, and fodder yield and quality.This module was also designed to determine the correlation and trade-offs between grain and fodder yield and quality. Regression analysis would be conducted for maize varieties of high feed potential to examine the relationship between grain yield and fodder yield and quality, and the stability of these parameters over time across regions. The module would also seek to develop criteria and indices for selecting maize genotypes based on the provision of grain and fodder. The information obtained could then provide the basis for developing criteria for selecting crop genotypes that satisfy food and residue demands of an array of farming system needs.Currently, some 60 varieties have been screened in terms of forage quality and quantity, to give an initial idea of the genetic variability that inheres in current varieties and to indicate possibilities for exploiting this variability in more formal selection trials (and possibly breeding trials) in the future, if this is warranted. All varieties were screened at low levels of fertiliser applications that might be expected in a farmer's field and at recommended rates.This component was designed to evaluate alternative forages that substitute and/or complement the use of crop residues as feed and soil amendment. The ex ante impact assessment would indicate which maize-related strategies needed to be investigated in on-farm experiments conducted at several locations over consecutive cropping seasons. Through partnerships with other existing projects, this activity would involve conducting on-farm and on-station field trials designed specifically to evaluate the feasibility of introducing alternative 'best-bet' forages (particularly grain legumes and fodder shrubs/trees) into cropping systems for replacing/enhancing crop residue use as livestock feed and soil amendments. Various cropping patterns with alternative crops/forages would be tested and compared with local cropping patterns and crop residue and land management practices.Quantification and evaluation would also be done of the long-term benefits and trade-offs in livestock and soil productivity when maize stover is grazed, harvested for stall-feeding or used for mulching. This would involve analyses of existing information and conducting appropriate trials to fill in information gaps on responses of livestock to feeding, and soils to application of maize stover. Livestock (live weight changes, milk yield), crop (grain and crop residue yield) and soil (runoff, soil organic matter and nutrient levels) responses would then be compared with those obtained using prevalent local practices. This component was envisaged as the integration of all project activities, where scenarios are investigated and 'best-bet' options are identified that can then enter testing in farmers' fields. This report describes the ex ante impact assessment work carried out over a number of months in 1999, 2000 and 2001 as a preliminary informationgenerating activity for this broader research project. As such, it combines most of Component 1 above with elements of Component 2, in that the analysis makes use of crop and livestock simulation models to assess the productivity impacts associated with particular system changes. Maize as Food, Feed and Fertiliser in Intensifying Crop-livestock Systems in East and Southern Africa 2 Technologies for the effective management of multiple uses of the maize crop This section reviews, in general terms, some of the issues and options relating to the adoption and impact of improved management practices for maize. At the broadest level, the SLP maize project's objective was to ensure that such management practices would contribute to more productive crop-livestock farming through higher and more stable biomass (grain and fodder) yields, while protecting the natural resource base. Accordingly, the technologies are discussed under three main category headings:• Technologies directed at improving the primary productivity of the maize crop • Technologies for promoting feed use of components of the maize crop and • Technologies for managing the soil resource for improved long-term performance.Necessarily, impacts directed at one of these aspects may have consequences for another. Where such interactions are likely to occur, attention is drawn to them.2.1 Technologies directed at improving the primary productivity of the maize cropSince the advent of the green revolution in the 1960s, the various approaches to genetic improvement have been seen as a reliable route to improved primary production in the farming systems of the tropics. On a global scale, many successes have been recorded; although uptake and/or impact have sometimes been compromised where changes to the production system, required to capitalise on the introduction of new germplasm, have been incompatible with farmers' circumstances or other objectives. As a result of the complexities governing farmers' responses to new technologies and the impacts of government policies, adoption has been widespread in some areas but patchy in others, and the countries of East and southern Africa (ESA) are no exception to this (Table 1).Some observations of the course of maize adoption in Malawi provide an informative example (see Box 1) of the complexities associated with the uptake of this, or indeed any, new technology at the farm level.Early introduced hybrid varieties proved unsuitable for home processing and storage. Farmers prefer flinty (hard) varieties with lower on-farm storage and home processing losses than dent (soft) varieties. Grain marketing authorities set uniform prices , regardless of grain type. As a result, farmers came to regard hybrid maize asa a cash crop whilst the local, flint varieties contiuned to be grown for home consumption. In addition, after storage and processing losses, many hybrids were found to be less acceptable for local use, when grown at low imput levels, than traditional varieties (Smale 1991).Since the early 1980s, use of hybrid seed and subsidised fertliser in the country has increased. However, average yields have remained static. Whilst areas planted to hybrides have increased, the negative effects of declining soil fertility would appear to have increased, the negative effects of declining soil fertlity would appear to have negated any beneficial impacts of genetic improvement. The yield potential of hybrids for Malawi exceeds 10 t/ha but, on impoverished soils, local varieties, composites and hybrids all give similar yields. this is likely to have serious repercussion from a livelihoods perspective. Kumwenda et al. (1996) have suggested that amongst smallholders using hybrid seed and applying the recommended amount of fertiliser to harvest 3t/ha of grain, over 40% fail to recover fertiliser costs, because of local conditions.Whilst the potential of varietal selection and hybridisation is great, their application without due consideration of the wider contexts of market pulls, the nature of the farming systems targeted and varied and changing environmental conditions may lead to disappointing results. In Malawi, where hybrid adoption has been particularly slow, the introduction of new hybrids, developed to take specific account of farmers' preferences, appears to be leading to too much wider acceptance (Smale et al. 1993). Varieties that are tolerant of a wider range of conditions may also be more acceptable. In Kenya, fertiliser use on maize has been greatest and spread most rapidly in the higher potential zones. Levels of fertiliser use are twice as high in the high-potential as compared to mediumpotential zones, and very low in marginal environments, reflecting levels of risk and returns (Hassan et al. 1998b). Clearly, suitable varieties for the drier areas should reflect a realisation that they may not always receive optimum fertiliser applications.Quantifying the achievable impacts of hybridisation or varietal selection on primary production is difficult because potential yields for improved varieties cannot always be realised in the field. Accordingly, an impact assessment such as this should be based on realistic evaluations of likely outcomes, given the range of factors operating. Furthermore, the timescales over which impacts may be judged, ex post, often exceed donor perspectives. In Kenya, for example, where over 70% of farmers have ultimately adopted hybrid and other improved varieties, the lag time between release and adoption by just 10% of farmers has varied from 4 years with a hybrid variety to 13 years for a composite variety (Hassan et al. 1998a). In such cases, the difference between success and failure may only be a pragmatic appreciation of the time required for a full and rounded maturity.The build-up of pest and weed infestations generally reduces yields of both maize grain and stover. Impacts may vary from relatively small economic losses, caused by competition with localised populations of weed species, to complete loss of economic yield, caused by plagues of insect pests or major invasions of weed species. The significance of weeds at the system level can be difficult to evaluate as a result of the different nature of their impacts in different system components. The case of Striga spp. in the region illustrates this (see Box 2).Box 2. Management and control of Striga spp.Striga spp. are pernicious weeds of arable crop land throughout East and Southern Africa. In addition to reducing grain yields, Striga infestations may also have an impact on foddr production and quality, although the nature of these impacts is not entirely clear. In one trial conducted in the northern Guinea savannah of Nigeria, Striga tended to reduce the edible forage to stem ratio and the dry matter digestibility of edible forage in late-maturing hybrid varieties tested but ot increase the dry matter digestibility of early maturing, open-pollinated varieties. In another group of open-pollinated hybrids, edible forage ot stem ratio was reduced in infested plants. Considerable variation was observed among hybrid varieties, which might be exploited as part of a selective breeding programme.The spread of Striga spp. is particularly favoured by low soil fertiltiy that, alos being widespread, tends to exacerbate their importance. A number of studies have considered the benefits of soil fertility improvements for Striga control. Imcorporating crop residues can assist in controlling Striga parasitism, particularly in a season following a relatively good harvest. Long-term studies have indicated that combinations of organic and inorganic sources of N are extremelty important in Striga control. Indeed, Odhiambo and Ransom (1995) suggested a strong additive effect between the two sources of N. However, Smaling et al. (1991) reported that application of mineral fertilisers or manures had no significant effect on Striga infestation and recommended that trap crops which allow Striga seed to germinate but not develop inot mature plants (e.g. sunflower, soya bean) should be used where the problem was particularly severe.'Weeds' growing on cropland are not always viewed as a problem because they can provide a valuable feed resource for livestock in intensive systems and systems under pressure. In Kenya, weeds are collected during land preparation and after first weeding for stall-fed animals (Getz and Onim 1993). In Ethiopia, farmers leave some weeds species in the field when weeding (Amaranthus, for instance, which is used for home brewing). Later in the season, leguminous weeds may be left to provide fodder for grazing animals after the crop and crop residue have been harvested (Franzel and van Houten 1992). Such weeds may even represent a source of cash income. Forage sellers in Tanzania include weeds such as Commelina in the fodder (Massawe et al. 1998). In most cases, the gardeners' definition of a weed as 'a plant that is growing in the wrong place' may be more appropriate than a rigid view.The importance of impacts on the quality of the forage as animal feed may be more ambiguous. In many systems where fodder is in short supply, variation in quantity is much more important than variation in quality for determining nutrient supplies to the animal. The balance between these two factors should always be borne in mind for the system under study when evaluating the significance of variation in feed quality. Nevertheless, it is certainly possible for crop pests and diseases to exert direct influences on the performance of livestock consuming affected plant parts (in addition to influencing the productivity of the crop component of the system). Organic matter in sorghum leaves infected with fungal pathogens may be reduced by as much as 20% when compared with healthy material. Leaf infection may also lead to early senescence, compromising the quality of thinnings and residues (Julian et al. 1994). Animals in the system may, however, play a beneficial role in weed or disease management. The passage of infected material through the gut may reduce the viability or even destroy causative agents. However, this is not always the case. Some weed species require passage of the seed through an animal in order to prime them and when deposited in a manure heap-an ideal environment for their development-may initiate further weed infestations. In this situation, composting manure, possibly through the higher temperatures generated, is effective in reducing the viability of grass seeds that have survived the digestive system (Shayo 1998).A further complication is that control methods employed to minimise the impacts of one pest or disease may exacerbate the effects of another. In the maize-bean intercrops common in East Africa, incorporation of stover into soil can reduce the severity of Rhizoctonia root rot of beans by suppressing the growth of the causative agent, Rhizoctonia solani (Lewis and Papavivas 1974). However, where maize stem borer is a problem, re-incorporated stover can form a reservoir for re-infection during the next growing season. Consumption of affected stover by livestock leads to the destruction of the stem borer. Moreover, the beneficial effects on Rhizoctonia populations are achieved by the immobilisation of soil nutrients, which in itself may adversely affect production levels.An assessment of the impacts of pest and weed management would appear to present some complex difficulties in smallholder maize production. An intervention that may be beneficial and acceptable to the farmer in one situation may be wholly inappropriate in another because of a different balance of objectives. This is likely to be a particular issue where livestock are closely associated with the maize crop.In the countries of ESA, maize forage for ruminant feeding may be derived from green stover (thinnings, leaf strippings, plant tops or the entire green plant after the immature cob has been picked for roasting) or dry stover from the mature plant after grain harvest. Maize bran is used primarily in non-ruminant diets but also to supplement forage-based ruminant rations. Grain and maize germ are predominantly used in commercially produced non-ruminant feeds. Only about 5% of maize grain in ESA is fed to livestock, with the exception of RSA, where the figure approaches 50%. Under severe human population pressure, such as in the highlands of Ethiopia, cereal straws and stover may also be used for bedding, fuel and house construction (Zinash and Seyoum 1991). The primary connection between a maize crop and its associated livestock is the conversion of a residue with generally no market value into a tradable commodity (meat, milk, draft and in some cases manure). Dry maize stover, available after grain harvest, is most widely used. The contribution of maize stover to the rations of ruminant livestock varies widely, depending upon human population density, the type of livestock, the management system, market access and climate.The maize plant can provide green fodder in various ways but collection is generally labour intensive and care must be taken to avoid damage to the developing plant.The practice of dense planting and subsequent thinning to obtain green maize forage is widespread in ESA. Dense planting followed by thinning at around 3 weeks when the first weeding occurs is a common practice that can compensate for patchy germination and provide a yield of green fodder. Maize yields generally exhibit an inverse U-shaped response to planting density; plant barrenness increases with density to a point where it negates any increase in yield brought about by an increase in the number of plants (Holliday 1960). Delayed thinning increases plant competition and reduces ultimate grain yields (Eddowes 1969;Francis et al. 1978;Schoper et al. 1982). However, Duncan (1972) and Fischer and Javed (1986) have suggested that, under favourable growing conditions, a greater initial density that increases leaf area index early in the growing season may increase total dry matter (DM) yields. Hence the trade-off between grain and fodder will depend critically on the plant population at each stage of the growing season. Farmers' willingness to make that trade-off will be related to the relative prices of maize grain and fodder. Furthermore this trade-off is likely to be a function of genotype, because genotype by planting density interactions have been widely observed in maize (Bunting 1973;Francis et al. 1978).Forage from immature plant thinnings contains up to six times the crude protein (CP) content of dry stover, and can produce up to 4 t DM/ha of high-quality feed without significantly affecting the final grain yield (Onim et al. 1991;Methu 1998). Inevitably, however, there are costs in both the long-term (soil nutrient depletion) and short-term (additional labour requirements).Thinning has been reported in western Kenya by a dual-purpose (meat and milk) goat project, in the highlands of Kenya, in Kilimanjaro region of Tanzania and in the Ethiopian highlands on smallholder dairy farms (Franzel and van Houten 1992;Getz and Onim 1993;Shirima 1994). Intensive dairy systems are most common in areas with high population density, good market access and small farm size resulting in a shortage of feed. In these systems, the cost of labour for thinning is most likely to be covered by additional milk production and sales representing a viable option for fodder supply (Lukuyu 2000).In addition, the adoption of multiple-tillering maize varieties may generate green forage for livestock feeding from the maize crop. Semenye et al. (1991) reported that Maseno Double Cobber, a variety selected from local material that produces two cobs per plant, also produced multiple tillers resulting in higher forage yield per plant. The International Maize and Wheat Improvement Centre (CIMMYT) maize programme in Ethiopia included a multiple-tillering variety in the early screening stages of evaluation that out-yielded other varieties. A few of the local landraces in Ethiopia have also been observed to produce multiple tillers but there is no information on how extensively this characteristic occurs or how farmers view it. Multiple tillering would appear to have definite advantages for forage production, and may merit further investigation.Other methods of obtaining green stover are generally less detrimental to soil fertility but produce less feed. Leaf stripping involves removing the bottom leaves from the plant sequentially over a period of time. Topping is the harvest of plant tops at silking stage. Both of these practices are labour intensive and less commonly observed although they produce better quality feed than harvesting dry stover without significantly affecting grain yield (Otieno et al. 1992;Shirima 1994). Semenye et al. (1991) found similar results for leaf stripping in western Kenya. Maize as Food, Feed and Fertiliser in Intensifying Crop-livestock Systems in East and Southern Africa Maize components were estimated to provide 30% of annual feed resources for livestock, based on studies in four villages in the Kilimanjaro region (Shirima 1994). However, the time taken to collect 1 t of leaf strippings was 106 hrs compared to 24 hrs for the same quantity of toppings and 58 hrs for strippings plus toppings from a pure maize stand. It is likely that collection would be even more labour intensive in intercropped maize.The importance of dry stover in the diet varies from 15% in the annual ruminant diet in the low veld of Zimbabwe, where extensive grazing lands provide the bulk of dry season feed (Steinfeld 1988), to almost 40% in cut-and-carry dairy systems in the Tanzanian highlands (Shirima 1994). Sandford (1989) showed that the importance of cereal stover varies with season in semi-arid West Africa, where an overall annual contribution from sorghum stover of as low as 16% climbed, in some areas, to 80% during the peak of the dry season. In Niger, up to 35% of the value of the cereal crop arose from stover use by livestock. Part of the difference between West and East Africa can be explained by rainfall patterns; two growing seasons per year in the East African highlands produce more stover than a single growing season in West Africa. Other factors include the ratio of livestock, especially cattle, to maize cultivation in a given area.Dry, mature maize stover, with low nitrogen (N) and digestible organic matter contents is, at best, a maintenance feed. Nevertheless, it is of considerable value during dry seasons when forage of any kind may be in short supply. In this situation, particularly in more extensive systems, poor quality natural grazing and dry, mature stover may comprise the major part of a ration. This is usually accompanied by a loss of live weight (see Osbourn 1976, for example) that provides the nitrogen and energy required for essential metabolic processes and a limited level of milk production in lactating animals. Subsequent compensatory growth may result in more effective use of the improved feed availability and quality that accompanies the onset of rainy periods (Topps 1976). For this reason, it may be more effective to focus supplementary feeding during dry seasons on animals that are pregnant or in early lactation as re-conception rates are related to body condition at critical times (Lamond 1970). In more intensive, commercially oriented systems, such as smallholder dairy systems, stover is likely to remain a small part of the diet supplemented with better quality material such as grasses, legume hays and agro-industrial by-products to ensure that an economically viable level of production may be sustained. It is noteworthy that the most developed dairy systems in Africa are found in highland regions with a bimodal rainfall distribution, where dry seasons are relatively short and green forage is available for longer periods during the year. Large-scale, commercial dairy systems in highland areas with a unimodal rainfall pattern, such as those in Zimbabwe, tend to rely on irrigation to produce high-quality forage during the dry season. In these situations it would appear unrealistic for poorer quality dry stover to form the basis of an optimised, productive feeding system.Even when substantial quantities of forage are available, their utility for animal feeding may be limited by a lack of access. For example, livestock may be tethered during rainy seasons to prevent damage to growing maize and other crops, restricting access to fodder. In this situation, the adverse impacts of limited feed consumption may be further compounded by increased helminth challenge, reducing growth rates and reproductive performance and increasing offspring mortality (Hendy and Carles 1993). In this study, production parameters improved after crop harvest when animals were released and had access to crop residues. In situations where a range of outputs other than meat and milk are valued, animal populations may be larger relative to the available feed resources. Interventions targeted on maximising meat and milk production via improved feeding are unlikely to reflect the needs of the farmer in these situations.The consumption of available stover by livestock is dependant upon maturity, variety and the amount on offer. Animals are selective and, given the opportunity, will choose the most palatable and nutritious plant parts first. Therefore, increasing the amount on offer allows a greater degree of selection as well as increased levels of intake. With dairy cattle fed maize stover, increased offer rates have been shown to result in higher milk production (Methu 1998). Although grazing cattle in Nigeria consume almost all the leaves from cereal stover left in the field, up to 50% of the less palatable stems are rejected when availability is high (van Raay and de Leeuw 1971). Chopping stover generally prevents cattle from selecting in this way but small ruminants, with a more flexible jaw and smaller mouthparts, are still able to choose leaf material over stems. Similarly, stall-fed animals offered excess feed tend to select leaf over stem (Osafo et al. 1994a, b). Zemmelink (1986) and Zemmelink et al. (1992) have discussed the importance of offer rates in determining the degree of selection possible and resultant animal performance. The level of excess feed required for maximum intake varies with the nature of the feed from 15 to over 40%. If either more or less than the optimum level of excess feed is available, production per unit of feed decreases. For tropical rations, Zemmelink et al. (1992) asserted that quality should be determined on the basis of voluntary intake of metabolisable energy (ME), incorporating digestibility as well as dietary CP levels. This is in contrast to the emphasis placed on degradable protein after meeting ME requirements, the basis on which ruminant diets for high-producing animals in the developed world are evaluated (e.g. CSIRO 1990;AFRC 1993).Most reports indicate that supplementation of a stover-based diet with legume forage increases total intake, while at the same time reducing stover intake (i.e. substitution occurs). This will result in higher nutrient intake and improved performance. However, effects on intake are variable, making their prediction problematic. For example, in a trial with sheep, using mature maize stover fed ad libitum and legume supplements, Smith et al. (1990) found that stover intake tended to decrease (although not significantly) when supplemented with lablab or pigeon-pea, and was unchanged by cowpea supplementation and rose when cottonseed cake was the supplement. In all cases, total intake increased as the level of legume supplementation increased. Even where effects are consistent, Maize as Food, Feed and Fertiliser in Intensifying Crop-livestock Systems in East and Southern Africa there are limitations in the published data. Most of the trials reported to date have employed insufficient treatments for establishing the response curves that would be needed to generalise the effects of different levels of a dietary component or even the basal ration for stall feeding. Nevertheless, it is clear that intake will increase as the quality of a forage improves. Thus, the intake of leafy green stover is higher than that of dry mature stover, and intake of dry stover will decrease as the leaf-to-stem ratio falls. Indeed, this type of effect has been reported in the field with livestock grazing stover in situ (Powell and Bayer 1985).2.6 Technologies for managing the soil resource for improved long-term performance Plant growth requires access to a range of nutrients in the rooting zone of the soil.The two most limiting nutrients for food production in Africa are N and phosphorus (P), in that order. A series of trials in Kisii, Kenya, found N deficiencies in 57% of the sites studied and P deficiencies in 26% (KARI 1994). At the level of soil solution, the nutrient balance represents plant-nutrient availability (Smaling et al. 1997).In weathered, tropical soils, where inorganic nutrient reserves are limited, organically held nutrients are essential for plant nutrition and nitrogen availability. Accordingly, the role of soil organic matter (SOM) in maintaining soil fertility cannot be emphasised too strongly. Soil organic matter levels are determined by the balance between the rate of synthesis of SOM (from plant residues and other dead organic materials) and its decomposition. Conversion of land to agriculture generally leads to a decline in the equilibrium level of SOM because of lower rates of organic matter input, and an increase in decomposition brought about by tillage operations. In cropped land, the main organic residues available for decomposition are roots remaining from harvested crops, aboveground crop residues that remain after harvest or grazing and, in some areas, manure. The retention of organic residues offers a means of maintaining or re-establishing higher levels of SOM equilibrium via the decomposition of crop residues (Ingrams and Swift 1990).Temperature is generally favourable to high rates of organic material decomposition and soil moisture is, therefore, usually the controlling physical factor. Thus, with seasonal tropical rainfall, wetting of organic material in or on soil is usually the trigger initiating rapid decomposition. Soil microbes effect mineralisation of organic material. Part of the organic material in the soil will decompose within 1 to 5 years, some between 5 and 20 years and the more intractable fraction may take between 20 and 100 years. A relatively stable flow of nutrients from the soil to plants growing in it requires a balance between stable SOM fractions that slowly decompose, and more readily broken down material. In Kisii, Kenya, Smaling (1993) calculated that the most common soil type contained 5 t N/ha that, at an annual mineralisation rate of 3%, is able to supply 150 kg N/ha per year.Phosphorus deficiency is a serious constraint to large areas of cropland in subhumid and semi-arid Africa. Phosphorus dynamics are complex, involving both chemical and biological processes and the long-term effects of sorption and desorption (see review by Sanchez et al. 1997). The low concentration and solubility of P in soils frequently make it a limiting nutrient. Phosphorus inputs consist primarily of organic sources (biomass, composts and manures) gathered outside the field. However, these are usually insufficient to meet crop requirements. A realistic application rate of organic material of 4 t DM/ha per year supplies only the 18 kg P needed by a 4-t/ha crop of maize grain. An exception may be manure from stall-fed dairy cattle offered mineral licks with a high P content (Lekasi et al. 1999).Based on data at national level, Stoorvogel and Smaling (1990) modelled the nutrient balance for countries in sub-Saharan Africa (SSA). For all major nutrients, the overall balance was negative despite the use of inorganic fertiliser (Table 2). Especially severe nutrient depletion was predicted for densely populated countries (Ethiopia, Kenya, Malawi, Rwanda), whilst low or zero depletion rates were estimated for sparsely populated semi-arid environments (Botswana). In Ethiopia, where the rate of depletion was particularly severe, soil erosion was deemed a prime cause. Source: Stoorvogel and Smaling (1990).On undegraded soils, any additional yields realised through the adoption of hybrids with limited application of inorganic fertiliser-the system generally favoured by farmers-simply leads to faster soil nutrient depletion. Continuous maize cultivation can result in declining yields and impoverished soil, although it may take several years for the impact on crop and soil to become apparent. Smaling (1993) presents data suggesting that, on a variety of soils in Kenya and Tanzania, impacts on production levels are only apparent with continuous cropping for more than four years.Farmers have long practised intercropping or, more generally, multiple cropping in a variety of forms. Indeed, it can sometimes appear that researchers are struggling to catch up with them. Ngwira et al. (1990) provide an example of the multiplicity of intercropping systems found in one country, Malawi (Table 3). None of the intercrops shown in Table 3 are grown specifically for forage, reflecting the overwhelming priority placed on the provision of human food. Quoting data from 1970, Ngwira et al. (1990) have indicated that, in Malawi, over 90% of maize, groundnuts, pulses, cassava, millet, sorghum and potato was grown as intercrops and that, in 40% of the cases, crop combinations included maize. Maize-pigeon-pea-cowpea -grams-sorghum-groundnuts Maize-pigeon-pea-grams -ground bean-groundnuts Source: Ngwira et al. (1990).Intercropping of cereals and legumes is common in many parts of Africa. The major practical advantage of intercropping is often an improvement in land equivalent ratios (LER). 1 Although yields of maize in intercropping are lower than those from a sole crop, total production of an intercropped plot is higher than if the components of the intercrop had been grown separately. Mureithi et al. (1996) at the Kenyan Coast found that the LER for a maize-cowpea intercrop exceeded one (the maximum possible value for a sole crop) in four out of the five seasons studied. However, mean grain yields for maize under intercropping were 51% less and for cowpea 12% less than in the respective sole crops. Furthermore, maize stover yield was 14% lower under intercropping, although the additional legume stover may more than compensate for this because of its higher nutritive value. However, a number of potential disadvantages are associated with intercropping, some of which are likely to be expressed in the longer term. Although legumes fix 1 The LER is the relative land area under sole crops that is required to produce the yields achieved by inter-or multicropping, all other things (management, inputs etc) being reasonably equal; this is in effect an index of biological efficiency related to growing more than one crop in a specific environment (Reddy 1990). atmospheric N, this does not necessarily benefit the companion cereal crop or improve soil fertility for future seasons. In maize-cowpea mixtures, Ofori et al. (1987) found that, in the absence of fertiliser, the amount of N fixed in aboveground biomass of cowpea was no greater than the amount of N removed in cowpea grain. Thus the removal of maize grain, in addition to cowpea, resulted in a net outflow of N from the plot. The quantity of N fixed varies amongst different leguminous crops. For example, soya bean fixes more N than cowpea. Nonetheless, Eaglesham et al. (1982) showed that soya bean caused a greater net loss of soil N than did cowpea because a larger fraction of soya bean N was sequestered in the grain. Giller et al. (1997) have summarised the impact of legumes as sole and intercrops on nitrogen fertility. Their data suggest a tendency for grain legumes to reduce soil N, and for forage legumes to increase it. As a result, it is rare for a legume, intercropped with maize, to be grown primarily as a soil improver. Instead it will be grown for grain or forage and when these products are removed there will be a loss of N from the system (Ingram and Swift 1990). The budgets of nutrients other than N will not be augmented by fixation and consequently a greater deficit of these might be expected in an intercropping system to which inorganic fertiliser additions have not been made. As a result, most researchers agree that intercropped mixtures extract more nutrients from the soil than does a single stand (e.g. Agboola 1980).Maize can also be intercropped with herbaceous legumes and livestock are often offered a mixture of forage, including different plant species, in their rations. The palatability of leguminous material is usually higher than that of dry stover but inevitably some will be wasted by trampling, or remain as feed refusals forming part of the manure-compost mixture that will be returned to the field. High-quality organic matter (e.g. from herbaceous, grain or forage legumes) will decompose rapidly in soil and provide an immediate boost to plant growth. The decomposition rate of organic material from tree and shrub legumes depends on the levels of polyphenolics in the material. Low-quality organic matter (e.g. cereal stovers) can adversely affect crop growth in the first season although benefits with regard to soil nutrient levels (Jones 1971) and crop yields (Tanner and Mugwira 1984) may become apparent in subsequent years.Although researchers have paid considerable attention to the development of improved legume technologies, their uptake has been slow in practice for a number of reasons. Labour shortages are an important constraint in many areas, irrespective of farm size. The phosphorus fertiliser needed to jump-start the system may be costly or unavailable and seeds may be difficult to obtain (Giller et al. 1994). Grain legumes would appear to have the fewest barriers to adoption and are widely grown for home consumption. However, they add little to SOM or soil N because most of the aboveground biomass is removed from the field, together with almost all of the N in grain. Species producing some grain with high biomass, such as pigeon-pea (Cajanus cajan) and dolichos (Lablab purpureus), might represent a useful compromise for combining adoption potential with improving soil fertility. Late-maturing pigeon-pea has its early growth reduced by competition Maize as Food, Feed and Fertiliser in Intensifying Crop-livestock Systems in East and Southern Africa with maize but compensates after the maize harvest to produce large quantities of biomass on residual moisture. The seed can be harvested for food and the leaf fall provides a significant contribution to N accumulation in the soil. However, pigeon-pea is attractive to grazing livestock and it is rarely practical to grow it where livestock roam freely after harvest (Sakala 1994). Cereal-legume rotations appear to offer greater prospects than intercropping to raise the yield of cereals (Natarajan and Shumba 1990). Maize-groundnut rotation studies conducted over a 5-year period in Zimbabwe showed that the inclusion of 1 year of groundnut almost doubled the maize crop in the following year (Waddington and Karigwindi 2001). However, the profitability of the rotation was less than for continuous maize with fertiliser and, in many situations, farmers may not be willing or able to sacrifice staple food production during the period needed to establish the system.A promising system in eastern Zambia that over 4000 farmers have adopted, involves the use of improved, short-term fallow with Sesbania sesban. The major plot-level impact of improved fallows reported by farmers in Zambia (as in western Kenya) is improved yields of maize (Kristjanson et al. 2002). Sesbania produces up to 6 t/ha per year of leafy biomass, providing an annual input to the system of about 120 kg N. When the fallow is cleared for planting, up to 7 t/ha of roots remain, which rapidly decompose, releasing nutrients into the soil. Sesbania helps to reduce Striga infestation, triggering germination during fallow periods and reducing seed bank levels during the subsequent maize crop. Over a 5-year period, following an initial 2-year fallow and without subsequent inorganic fertiliser additions, final-year yields of maize grain under improved fallow were 2.3 t/ha compared with less than 1 t/ha on fields that had been continuously cropped (Kwesiga and Beniest 1998). Total maize grain production over the 5-year period was 9.8 t/ha with the improved fallow compared with 7.6 t/ha under continuous cropping.On sloping land in Malawi, on which farmers produced local maize yielding about 0.5 t/ha, Sesbania, undersown in semi-flint hybrid maize and harvested prior to the following rainy season, produced sufficient N in biomass to reduce N requirements for the maize crop by two-thirds at the top and bottom of the slope, and by half in intermediate positions (Phiri et al. 1998). As a result of the shortterm Sesbania fallow, maize grain yields increased three-fold and stover yields two-fold compared with the controls.The stover fraction of the maize plant contains fewer nutrients than the grain. However, the removal of stover as fodder, construction material or fuel still represents a significant additional outflow of nutrients from the plot. Crop residues left on the soil surface can provide a physical barrier to soil movement, reducing soil erosion, as well as enhancing soil chemical and physical properties by providing a substrate for soil microbes. Improved soil structure also increases infiltration rates and reduces surface evaporation leading to greater water storage capacities (Powell and Unger 1997). However, over most of ESA, farmers do not have the option of leaving crop residues on the ground for soil protection because they are required for fodder, either being removed by farmers for dry season supplementary feed or consumed in situ by free roaming, grazing animals. In this respect, animals might be deemed to contribute to soil nutrient depletion although this may be alleviated partly by recycling of manure.Any crop residues not removed by farmers or eaten by livestock are either incorporated in the soil by insect and microbial activity or purposely burnt; the latter also resulting in losses of volatile nutrients. This strategy may be adopted, nevertheless, to facilitate manual cultivation and control pests and diseases. Excessive removal of crop residues can deplete SOM and nutrient reserves and increase the risk of soil erosion and degradation. The estimated maximum amount of N that could be returned when livestock graze crop residues equals the N returned in trampled fractions of crop residue, in manure (faeces and urine) and via rain and soil bacteria. Nitrogen balances probably approach equilibrium in some places where there is widespread application of fertiliser on cash crops and selective application of manure during cropping periods.Although the SOM of most cultivated soils is highly buffered and maintained at fairly constant (albeit low) levels, dramatic changes in crop residue management can disrupt the SOM equilibrium. With the adoption of high-yielding crop varieties, for example, the extraction of nutrients from soil will increase. If residues are returned to the soil, SOM and nutrient levels will improve (Jenkinson and Rayner 1977;Grove et al. 1986). Conversely, without the return of crop residues, SOM and available nutrient levels will decline (Powell and Hons 1992). Cereal stover, with a high carbon (C):N ratio, maintains an equilibrium between soil nutrient immobilisation and mineralisation processes. If either input or output alters, the equilibrium is disturbed. The effects of stover removal are more pronounced on coarser textured soils having lower SOM and available nutrient reserves and high SOM turnover rates. During the early SOM decay process, labile SOM fractions are the first to mineralise followed by the more stable SOM components (van Faassen and Smilde 1985). Because these are the fractions most strongly influenced by crop residue management strategies (Elliott and Papendick 1986), the impacts of changes in crop residue management are generally most pronounced in the short-term. Nandwa (1995) studied the effect of surface placement (mulching) or incorporation in soil at various depths on the rate of decomposition of maize stover at semi-humid and semi-arid sites in Kenya. The half-life for stover decomposition was significantly shorter with deep incorporation and was longer at the semi-humid site. With sufficient soil moisture, application of N fertiliser increased the rate of decomposition but there was no effect in dry soil. Methu (1998) used urea-treated stover, simulating the situation in which stover or feed refusals form the bedding of stall-fed cattle that had been exposed to urinary and faecal contamination. Urea treatment significantly increased the decomposition rate for stover incorporated in soil. As might be expected, decomposition rates for stover components were, in descending order, leaf -husk -sheath -whole stover -stem.A major benefit of the integration of livestock and crops into mixed farming systems (based on maize or otherwise) is the potential for improving nutrient conservation, importation and availability via the feed (including residues)animal -manure pathway. The importance of manure applications to the maize crop varies amongst farming systems in ESA and depends on the degree of intensification, cattle numbers in relation to areas of maize planted and the priority given to maize within the cropping system (Table 4). In intensive dairy systems in which stall-feeding is practised, manure is relatively easy to collect. In extensive systems, cattle may be kraaled (fenced in) at night, so that a proportion of total faecal output is readily available but faeces deposited while grazing are only of agricultural value when animals are on fallow or grazing crop residues in situ. In the central region of Malawi, maize is a priority crop for home consumption but there are few cattle and hence little manure. While crops are growing, livestock are tethered on roadside grass or uncultivated land and manure is not collected. In the dairy systems of the Kenya highlands, maize is grown on about one-third of the farm for subsistence but also provides a valuable source of cattle feed. Manure is usually the principal source for replenishing soil P and N, originating from boughtin supplementary feed and minerals and, to a lesser extent, the forage fed.  Unfortunately, many small-scale farmers own insufficient livestock to apply the necessary quantities of manure to obtain a noticeable response. The approach in many systems is therefore to concentrate manure use on smaller areas of land, on which the most valuable crops are grown, and to practice crop rotation. This is generally inadequate for maintaining the nutrient balance of the farm as a whole. Nevertheless, in a comparison of sites in Kenya where manure is derived from animal grazing on communal lands, Smaling et al. (1997) showed that the nutrients imported onto the farms helped to raise balances closer to zero. Similarly, with stall-fed dairy animals, commercial concentrates, mineral supplements or any other imported feeds also form a significant inward conduit for nutrients improving overall nutrient balances. In drier areas of southern Zimbabwe, farming is more extensive but each household owns fewer cattle. Thus, although more land is cultivated, less manure is produced, application rates are lower and less land is manured. In these areas, areas manured, areas of maize grown and maize yields are all positively correlated with the sizes of cattle holdings. Almost all of the manure produced is applied to maize with some preference displayed towards the fields closest to the kraal where transport costs are lower (Steinfeld 1988). Although manure by itself is therefore usually insufficient, farmers in drier areas face the risk of considerable financial losses if they choose to purchase and apply inorganic fertiliser. It may be possible to reduce this risk by using combinations of organic and inorganic amendments. The data of Raussen (1998) illustrate how the addition of organic material such as manure can raise SOM levels and improve fertiliser use efficiency (Table 5). Source: Raussen (1998).1. Top dressing with NPK 10:20:10 at 100 kg/ha, plus 100 kg urea/ha.Losses of manure N through ammonia volatilisation may be significant. These losses can occur soon after excretion of dung and urine, during storage and after application to the field. Ammonia volatilisation is affected by manure storage conditions-pH, temperature, and moisture levels-as well as the initial manure quality. Ammonia losses decrease as the manure C:N ratio increases. Up to 40% of excreted N can be lost from manure alone but incorporation in soil reduces losses by half (Giddens and Rao 1975;Kirchmann 1985). However, in Zimbabwe, Murwira et al. (1995) reported insignificant losses after field application, possibly because of the low N content of the manure found in communal lands. This contrasted with up to 6% loss of total N through ammonia volatilisation during storage.Over the past three decades, the main thrust of extension messages relating to soil fertility has been focused on the promotion of fertiliser use. However, application rates by smallholder farmers have been limited by availability, cost and the risks of financial loss. In many countries, the option of using organic material instead of, or as well as, inorganic fertiliser is very limited. In Malawi, for example, over most of the country, cattle populations are less than 0.25 cattle per hectare of cropped land (Coote et al. 1998). Even zero-grazed cattle in Kenya produce only an estimated 1 to 1.5 t manure DM per animal per year (Strobel 1987). At these levels (assuming no post-voiding losses), two animals would be required to supply sufficient manure nutrients for a maize crop yielding 2 t/ha if the manure is of high quality, or eight animals if the quality is low. Swift et al. (1989) calculated that 96 kg manure N, equivalent to 10 t of low quality manure, would be needed to maintain 2-t maize yields in Zimbabwe, equivalent to a stocking rate of 15 to 20 tropical livestock units (TLU) per square kilometre of miombo woodland grazing land. In semi-arid West Africa, Sandford (1989) calculated that to maintain a yield of 1 t/ha of maize, a farmer would need manure from one cow grazing on 23 ha of rangeland in high-potential areas, and 64 ha of low-potential rangeland. In the dry land area of Machakos, Kenya, Probert et al. (1995) calculated that the associated animal populations could provide around 2.5 t/ha of cropland per year. In this area, however, farmers were concentrating manure applications on just a few fields, resulting in application rates of up to 38 t/ha. Clearly, inadequacy of manure supplies is a widespread problem resulting in farmers applying a wide variety of strategies.Increasing the N content of the manure available for application might result in greater benefits when it is actually applied. A possible approach lies in the improved capture of urine together with faeces. This may be achieved through the effective use of bedding either in the form of feed refusals (e.g. dairy farmers in the highlands of Africa, stall-fed goats in the Republic of Benin) or especially collected material (forest litter in Nepal). This allows a high degree of spoilage and wastage of the feed but eventually a substantial amount of manure-compost may become available (Reynolds et al. 1995). In all of these systems, urine is trapped by the plant material, so that the resultant mixture contains faeces, urine and plant residues. Recent work in Kenya by Lekasi et al. (1999) suggests that there is a higher N content in manure-compost made solely from faeces and feed refusals than when urine is included in the mixture. In the absence of feed refusals, manure, either from faeces alone or faeces plus urine, produced lower quality composts. A maize fertilisation trial confirmed this conclusion. The application of all manure-composts improved grain yields when compared with unfertilised controls but manure-compost from faeces plus feed refusals produced the highest yields of both grain and stover. At the higher fertility site, manure from one cow would produce an incremental yield of over 400 kg grain and almost 500 kg stover. The practical implications of this are that faeces should be collected from the stall and piled on a compost heap to which feed refusals would be added directly. Urine can then be drained from the stall into a pit and removed periodically for separate application to the crop. The additional grain produced by this technique would have a significant beneficial effect on household food security, since an average Kenyan family of six people in the highlands are known to consume around 1 kg of maize flour per day (Staal et al. 1998).Maize dominates the cropping system for many small-scale and subsistence farmers in ESA. The crop is grown throughout the region, even under suboptimal conditions. Suitability for maize production is determined mainly by the length of the growing season that in turn is determined by the amount of rainfall and its distribution and temperature. In southern Africa, only a single crop is possible each year but in Kenya, Uganda and parts of Tanzania, two cropping seasons are possible as a result of the bimodal rainfall patterns found in these countries. Maize is generally grown in the areas of high-to medium-production potential that, because of their ecological and geographic characteristics, have the potential to be major food producing areas for Africa.Maize is highly important as a staple throughout the region, particularly in Malawi and Zambia, where it contributes two-thirds of the calorie supply to families, and in Zimbabwe and Kenya, where it contributes over 40% (Table 6). The area under maize in Uganda doubled from 1980 to 1996 but its contribution to the human diet is still much lower than elsewhere in the region. In ESA as a whole, the human population is growing at around 2.7% annually, resulting (locally at least) in increasing pressure on land and food supplies. In 1996, it may be estimated that about 80 million people received 25% or more of their calories from maize, so it is important not to underestimate the need for improvements in maize production that will keep pace with this population growth.Maize is an efficient converter of carbon dioxide and water to carbohydrate, producing more calories per unit area than any common alternative. However, in Malawi for example, where the crop covers 70% of the cultivated land (90% in the south of the country), farm sizes are less than 1 ha. In Zomba, Thyolo and Blantyre regions, with population densities of 214 people per square kilometre, maize yields are 1 t/ha or less. Over 60% of households in the country are net purchasers of maize, their staple food, for several months each year. Increases in overall production at national level over the past 30 years have arisen from expansion of maize area, rather than from improved yields (MacColl 1989). As mean farm size has fallen, reflecting population growth and demand for land, subsistence production per household has declined. Clearly, there is a great deal of scope for improving the contribution made by maize to rural livelihoods in the country. Increased pressure on land is accompanied by shorter fallow periods (Tanzania, Zambia, Mozambique) or by effective sedentarisation of agriculture (Malawi, Kenya, Zimbabwe). Governments in the region have given top priority to maize production, perceiving that maize in ESA is as important as is rice in Asia (Byerlee and Eicher 1997). Despite the importance of maize, African maize production is characterised by low yields and high variability compared to other parts of the world. Jayne et al. (1997) analysed maize production in communal lands of Zimbabwe from 1915to 1994. From 1915 to independence in 1980 the trend line showed yields per head declining by 18 kg each decade. The coefficients of In order to examine the production of maize and its significance in the farming systems of the region, case studies were conducted in five countries: Kenya -Thorne (1999a); Malawi -Reynolds (1999b); RSA -Le Roux (1999); Tanzania -Thorne (1999b) and Zimbabwe - Reynolds (1999b).This section briefly summarises the key information from the case studies (Tables 7-11; Figure 1 shows the general locations of the places mentioned in the tables). This information, and other data from the literature, have been used to generate the baseline data for the ex ante impact assessments described in subsequent sections.       Maize growing is generally confined to areas with mean annual rainfall of 500 mm or above. However, in areas such as the Eastern Cape Province of RSA, where this rainfall is less reliable (i.e. more variable), farmers may not consider maize growing to be a viable option. Indeed, crop failures are not uncommon-about 10-20% of maize crops in the region appear to be failing currently-and seem to be an increasing feature of maize production in ESA.The ranges in levels and distribution of precipitation found across the five countries under study can have considerable implications for maize production options with respect to the varieties used and the objectives addressed in growing them. For example, in areas exhibiting bimodal rainfall patterns, maize cropping may differ considerably between seasons. Lukuyu (2000) indicates that the shortseason crop in the Kenya highlands may be planted with little aim of harvesting a significant grain yield but is considered valuable for the forage that it can produce even when rainfall is limited. The varieties selected reflect these perceptions and objectives.An observation made in the RSA case study, that differences in the wealth of individual families could be explained largely by differences in off-farm earnings, probably holds true across much of the region. Indeed, landholding sizes have declined to such an extent in some areas of Kenya, Malawi and Zimbabwe that some alternative forms of income have become essential for survival. This problem appears to be associated particularly (and not surprisingly) with the more intensive farming systems practised in more densely populated areas (Kenya, Zimbabwe, parts of Malawi) and under particular forms of land tenure. There is some evidence of effective responses to declining landholding sizes (see below) but these would appear to be limited where land is subdivided as part of the process of inheritance and each generation faces a more acute and less easily soluble difficulty.In East Africa, and to a lesser extent southern Africa, the predominant maize cropping systems are based on intercrops, largely with beans but also with other crops such as coffee and banana. Maize yields realised continue to be well below potential yields in all five countries (the mean for RSA probably paints an unreliable picture because of the relatively widespread commercial growing for feed use). Much effort has been concentrated on yield improvements through genetic improvement and greater use of inorganic fertiliser. In the case of the former, it would appear that, in some of the countries at least (most notably Kenya and Zimbabwe), a state of diminishing returns is being reached. In the case of the latter, problems of access to inputs-because of infrastructural difficulties or simply Maize as Food, Feed and Fertiliser in Intensifying Crop-livestock Systems in East and Southern Africa resulting from a lack of funds in the farm household-often means that inorganic fertiliser use is not a viable option for poorer families. Some of the perceived successes of these technologies have been associated with input programmes that have subsidised fertiliser and seed costs and there has often been a degree of reversion to traditional varieties and production systems when the programmes have ceased to operate.Against a backdrop of generally declining soil fertility, and the practical difficulties of meeting the challenge that this poses solely through a reliance on inorganic fertilisers, interest in the role played by crop-livestock integration in soil fertility maintenance has increased considerably. Indeed, there are indications, in some areas at least (parts of Kenya and Tanzania), that progress has been made. Despite reductions in the sizes of household landholdings, more intensive and effective management of crop-livestock integration, particularly in manure and crop residue use, has led to improved production from the maize crop. Whether this might represent a model for some of the other countries studied, where livestock are often not closely integrated with cropping is, however, open to question. Despite maize's long-established role as the principal staple in ESA, there are indications from the case studies that some farmers are prioritising other crops. In much of Kenya and Tanzania, it would appear that manure from associated livestock is applied preferentially to cash crops. Where there is significant use of maize stover for livestock feeding, this practice represents an outflow of nutrients from maize land that may have particularly acute long-term consequences where maize and other landholdings are disaggregated (e.g. Kilimanjaro region, Tanzania). The widespread uptake of organised, smallholder dairying also appears to be affecting the extent of maize growing in some areas. Because of the difficulties in maintaining year-round fodder supplies, some farmers with cattle are replacing maize with planted forages. Clearly, potential interventions in the maize system must not ignore a perception that, for some circumstances, more radically altered cropping systems may be most appropriate for the farmer's needs.Crop-livestock integration is traditionally strong in Africa's mixed farming systems based around maize cultivation. In some situations (e.g. RSA) a perceived decline of this integration, particularly in the availability of draft power, has been viewed as compromising the maize production system. This is less significant in East Africa where hand cultivation is much more prevalent.The advent of the smallholder dairy cow with established marketing channels for milk has undoubtedly impacted on the maize production system as discussed above. Despite generally good markets for milk in the study countries, milk yields continue to be low (averaging only 9 litres per day in Kiambu, Kenya, for example). With crossbred cattle, there is no reason why, with proper feeding and management, such cows should not be averaging 15 to 20 litres per day. This difference in actual and potential production represents a considerable challenge for the feed resources that may be derived from the maize-based farming system. Whilst there are undoubtedly technologies that might be applied to the maize crop to increase its contribution to feed requirements, it seems unlikely that specialist dairying can be supported by resort to maize crop residues alone. In Kiambu, Kenya farmers have often been observed feeding unusually high levels of concentrate when milk prices are high in order to compensate for a lack of roughages. It would also appear that, in western Kenya, the Kenya highlands and parts of Tanzania, planted forages are slowly making inroads into land that might traditionally have been planted to maize. One farmer in Kakamega interviewed for the Kenya case study stated that financial returns from a given area of land increased in the ratio 1:2:4 in relation to the following progression: maize planted for sale ➔ napier grass planted for sale ➔ sale of milk produced from planted napier grass.A common theme from the case studies is the overwhelming reliance placed on traditional marketing channels. These include the village and town markets and the informal transfer of produce amongst neighbouring families. Some organised marketing channels do exist (Tanga Fresh Dairies and the Transkei Agricultural Marketing Board, for instance) but their effectiveness appears highly variable. Indeed, during the course of the Kenya case study, a major player in the marketing of milk in Kenya, the Kenya Co-operative Creameries, was declared bankrupt. Farmers interviewed during the Tanzania case study were particularly concerned about marketing via middlemen who were perceived as unreliable and making excessive charges for their services.A key element of this study lies in improving the matching of the available and potential technologies with the farming systems of ESA in order to generate beneficial impacts for the largest possible groups of impoverished people. This issue is examined in some detail in the modelling studies described in this section. It provides a brief overview of the relationships between the available technologies and their likely points of impact in the maize-livestock system. It also considers the scope of the likely impacts in terms of the distribution of the human population amongst a set of broader farming system types that accommodate maize-livestock production in ESA.Figure 2 summarises the major potential interventions and their main impact points within a generic maize-livestock system. Even at this level of aggregation, this diagram clearly illustrates the potential for individual technological innovations to interfere with one another (in both a negative or a self-reinforcing fashion). This may be expected to result in a range of trade-off situations that are likely to require more detailed analysis for individual circumstances. Table 12 represents an overview, based on the information presented in the preceding  2. +, ++, +++ represent varying degrees of positive impact;-,-,-varying degrees of negative impact; and ? uncertain, variable or unclear impact. 3. At critical times. 4. Although replacement of the maize crop will, of course, affect maize production at the farm level. 5. For maize but probably + overall. 6. In the longer term. literature review and on expert consultation, of the likely outcomes of the individual component interventions for the key productive elements of the system (arranged under the outcome groups; grain yield, feed yields, feed quality, livestock products and soil fertility). This format has been used as a unifying framework for presenting the findings of the modelling study (see below). Table 13 summarises the individual component variables of the production outcome groups that have been used to assess the interventions during the course of the modelling studies. The major crop-livestock production systems in the five countries under study have been characterised (indications of what 'typical' farms look like in the case study locations in each country can be found in Tables 7-11). In outline, this characterisation was conducted as follows:• A basic assessment of key system features was made, based on information contained within the country case studies summarised above as well as on literature review and expert opinion for each of the five countries. The individual characteristics used in this exercise included maize density, cattle density, human population density, farm sizes and the extent and nature of maize-livestock interactions (Tables 14 and 15). For practical reasons, it was assumed that the systems characterised would be homogeneous within countries but somewhat variable amongst them. • Geographic information system (GIS) analysis was used to estimate the areas and numbers of people in a range of population density classes that could be related to the characteristics of the four types of maize-livestock systems identified (Table 16). Information on the areas and populations associated with the different rainfall patterns found in ESA (unimodal or bimodal, with bimodal subdivided into seasons of equal or unequal length) was also generated (Table 17) using a GIS analysis based on the information contained within the CIMMYT African Country Almanac (Corbett et al. 1998).The information generated by these activities indicated that in Kenya, the largest areas with maize densities above 10% and up to the maximum of 40% are found in the human population density classes of above 250 people per square kilometre that correspond to small-scale, intensive farming systems. In Malawi, the highest maize densities of 60-80% and 80-100% dominate the human population density class of 30 to 100 people per square kilometre, corresponding to the medium-scale, semi-intensive farming system as defined above. This indicates that, based purely on maize density, the farming systems with the greatest potential to benefit from increased maize-livestock interactions would appear to be the smallscale intensive (SSI) farming systems in Kenya and the medium-scale semiintensive (MSSI) farming systems in Malawi.Taking cattle numbers into consideration, in particular dairy cattle, it was estimated that about 40% of the dairy cattle in Kenya are found in SSI farming systems. This feature, coupled with the information that these production systems have the highest maize densities, indicates a high potential for benefits from increased maize-livestock interaction. In the other three countries where this level of detail with respect to the distribution of cattle was unavailable, total cattle numbers were used to provide some insights into the potential for improved maize-livestock interaction in different production systems.It was apparent from the review of literature and the country case studies that not all of the technologies considered would be appropriate in every production system and country. For example, further promotion of improved varieties of maize is likely to be of only local relevance in the more intensive farming systems of Kenya, given that uptake of improved varieties has already been extensive and widespread in the country. Furthermore, the assessment of the relevance of individual technologies for different circumstances needs to be made not only on the basis of their impacts on the productive outputs targeted but also on their socio-economic feasibility and their side effects and knock-on impacts in other parts of the farming system. Accordingly, a set of matrices for technology-system matching was constructed (Tables 18 to 21). These have been designed to represent the technologies that are most likely to be relevant in each of the systems and countries under study to allow the pre-screening of the potential technologies  to be studied in the ex ante impact assessment. They also take into account the distribution of maize cultivation across the different systems in the five countries and, therefore, their relative importance in those countries. The information used in constructing these matrices has been derived from the literature and the country case studies.Based on the technology system matrices and the information in the literature, the following conclusions have been drawn regarding the potential interventions and the scope for their application.The current scope for more widespread application of straight genetic improvement approaches to enhancing primary productivity is limited because: • Where benefits are readily realisable (e.g. in the SSI systems in Kenya and Tanzania), uptake of these technologies has already been good. • The complementary inputs (e.g. fertilisers, soil and water management techniques) required to realise the benefits of genetic improvements are often not reliably available. • Climatic uncertainty and variability mean that investments in improved seed stocks may pose unacceptable risks for farmers. • The possibility exists that the feasibility of varietal selection on a large scale for characteristics related to fodder quality and yield could be compromised by the adverse impacts of improvements in these traits on yield stability (Bindiganavile Vivek, personal communication, 2002).As far as genetic improvement is concerned, the challenge lies in realising the potential of improved or hybrid varieties in farmers' fields. Data drawn from successive KARI reports by Lukuyu (2000) suggest that on-farm yields in different parts of East Africa are still only around 15 to 30% of achievable yields in different agro-ecological zones. It is not currently clear how the other inputs required for the expression of improved yield can be delivered widely in a way that is acceptable to farmers. As well as breeding programmes with the primary aim of increasing the size and stability of grain yields, some interest has been shown in genetic improvement of stover quality. It has been argued that small improvements in stover digestibility might result, overall, in quite large increases in the availability of livestock products. (The study of Kristjanson and Zerbini (1999) investigated the prospects for improving the digestibility of millet and sorghum residues in India and found attractive returns to that research investment.) Although data were not available to examine this option at the same level of detail as the other interventions, a brief case study is presented below, and the results used in the impact assessment proper. Maize as Food, Feed and Fertiliser in Intensifying Crop-livestock Systems in East and Southern Africa 1. RSA = Republic of South Africa.Benefits accrued from the pest and weed management techniques identified are more likely to arise as spin-offs from improved feed management techniques:• Where weed collection has a demonstrable impact on benefits from livestock production, it is also likely to affect weed populations. This will, of course, result ultimately in reduced availability of the feed resource so the long-term benefits of this method are probably questionable.• Passage of infected or infested material through livestock may be beneficial although, in some cases, application of manure derived from this material may spread pests or diseases more widely.The various management techniques aimed at extracting a useable feed resource from the green maize stover all offer some potential for improving the contribution of the maize crop to feed availability. Some potential, practical problems need to be examined, however:• Timing of availability. Often these feeds are available during periods of relative plenty. They do not, without preservative treatment, address the problems of dry season feed availability. • Quantity. In relation to the year-round feed requirements of associated livestock, quantities available may be relatively small.Dry maize stover is already the most widely used feed resource from the maize crop whether it is stored for use in the cut-and-carry feeding system associated with more intensive systems or grazed in situ under more extensive conditions. Potential interventions fall into three main areas:• Improving the integration of dry maize stover into the feeding system. This includes supplementation strategies and general rationing and year-round feed budgeting approaches. • Physical treatments, principally the impacts of chopping. This method is probably already widely used in the SSI system. Chemical treatments are not considered because their financial viability is always likely to be questionable; such treatment also usually requires large supplies of water, which may be difficult to obtain, and transport in the dry season. • Replacement with better fodder. The replacement of part or all of the maize crop with specialised fodder crops may actually represent the most attractive option in more intensive, specialised systems.Trials with improved fallow systems have shown promising results, at least in more extensive systems. However, these have usually been under experimental, or at least artificial, situations. The following issues would need to be addressed before these technologies could be widely promoted:• Input costs, particularly labour, are likely to be high. What are the implications of this for adoption? • There is little information on the range of options for fallow crops and how these might be varied in order to meet individual needs.Intercropping is already widely practised in many parts of ESA, particularly in more intensive production systems. However, there would appear to be some scope for the more widespread adoption of intercropping in more extensive systems, although the resulting increase in household labour requirements may pose a practical problem to adoption in some situations.The use of stover for mulching or soil amendment is likely to be a viable option only in the extensive and semi-intensive systems where its use in this way does not compete with its use as a feed resource.Effective manure management to conserve biomass and nutrient concentrations has considerable potential in intensive systems. Some relatively simple technologies are available that could be used to reduce overall nutrient losses or to improve the quality of manure-compost through augmentation with other biomass. Prioritising applications of available manure compost to particular components of the cropping system might represent an appropriate technology for more extensive systems but, for an exercise of this type, it would be difficult to appraise the extent to which farmers might wish to trade-off amongst different cropping enterprises. Again, the labour implications of transporting manure or compost to the field may pose a practical problem to adoption in some situations.Table 22 outlines the main interventions that would appear to offer potential if applied in the maize-livestock systems of ESA, their likely impacts and the systems and countries in which these are most likely to be realised. These interventions have been used as a basis for the more detailed impact assessments described in the next section of this report. 1. Systems: MSE = medium-scale extensive, MSI = medium-scale intensive, MSSI = medium-scale, semi-intensive, and SSI = small-scale intensive. Countries: KEN = Kenya, MWI = Malawi, RSA = Republic of South Africa, TZA = Tanzania, and ZIM = Zimbabwe.5 Ex ante impact assessment of technologies for improving the use of the maize crop for food and feed This impact assessment, based in part on a modelling approach, was carried out in order to assess the potential impacts of the technological interventions identified (see Table 22). These assessments of the biophysical impacts of the technologies were then used to generate input data for an economic surplus model, used to examine the likely economic outcomes of their adoption.The first part of the modelling study was designed to examine the biophysical impacts of a range of interventions in the maize livestock systems. Representative baseline data were estimated for the defining variables (see Table 13) in each of the four systems in all five countries. These values were derived mostly from the country case studies but, where reliable quantitative data were not available, extrapolations were used. For example, in some cases, values for milk offtake were calculated from the baseline data describing forage quality and availability using the Dairy Rationing System for the Tropics (DRASTIC) model (Thorne 1998).Having estimated the baseline data, adjustments were made to these in order to define the consequences, in terms of input variables, of the interventions that would be tested. It was judged that no one single biophysical model was available that could be used for this exercise because the impacts of each intervention on the soil, crop and livestock components of the maize-livestock system had to be considered. The following sections describe, in outline, the major components of the modelling approach taken during the course of the study. Some minor variations to the general procedures described were employed depending on the individual technology under consideration.The CERES maize and CROPGRO models (components of DSSAT 3.5, Tsuji et al. 1994) were used to evaluate grain and forage production from the maize crop and, for scenarios in which intercropping was envisaged, an associated bean intercrop. Key elements of the approach included manipulation of:• Spacing/planting densities to simulate (crudely) the effects of intercropping and pest/weed management practices; • Data describing initial soil conditions, used to examine possible impacts of improved fallows; and• Residue incorporation and quality (principally N concentrations) data to simulate different crop residue and manure/compost management schemes.Fodder quality was estimated from N contents predicted by the DSSAT models. In the case of the interventions based on dry maize stover use, these were based on the predicted values for percent N at harvest, while the green maize stover interventions used values for intermediate growth stages predicted by the models.DRASTIC (Thorne 1998) was used to examine the impacts of changes in forage quantity and quality on milk production under smallholder management. DRASTIC is a field tool that uses qualitative indicators of fodder quality to assess the outcomes of feeding interventions for milk production. In order to meet the requirements of the current, quantitative study, suitable combinations of qualitative indicators describing each intervention situation were derived as input data for DRASTIC. This was achieved by applying, in reverse, the algorithms used by DRASTIC to translate qualitative indicators into quantitative estimates to the fodder quality data generated by the simulation.The Allocation of Nitrogen in Organic Resources for Animals and Crops (ANORAC) model (Thorne and Cadisch 1998) incorporates a soil organic matter model based on CENTURY (Parton et al. 1987). This was used to indicate changes in soil nutrient and organic matter status following the application of the technological interventions. ANORAC also includes a component to predict impacts of bodyweight changes in livestock associated with a crop-livestock system. This was used, where possible, to derive values on which to base estimates of meat offtake in the more extensive systems.Where input data were not thought to be sufficiently reliable for a modelling approach to be applied, values were estimated from extrapolations of information contained within the literature.Tables 23 to 26 present baseline data for the four production systems in the five countries. Sources are indicated in the tables. Values were derived largely from the country studies commissioned by the project and, where these were inadequate, by recourse to the available literature. Various assumptions were made in deriving the baseline data in these tables:• Maize grain and intercrop yields were derived from estimated mean yields and the estimated proportion of land under intercropping in the target systems in each country. • In locations with a bimodal rainfall pattern, values for dry and wet seasons are aggregated. • Estimates of feed quantities relate to cattle of mean bodyweight 300 kg (metabolic bodyweight = 36.4 kg). • Estimates of meat offtake are based on predicted live weight changes and an estimated (variable) rate of offtake. • Estimates of manuring rates were based on available averages.  Franzel and van Houten (1992), 6. Heisey and Smale (1995), 7. Corbett et al (1998), 8. Other geographic information systems (GIS) data, 9. Thorne (1998), 10. Stoorvogel andSmaling (1990), and11. Thorne andCadisch (1998). 3. CP = crude protein, and DM = dry matter. 4. MBW = metabolic bodyweight of livestock. 5. OM = organic matter.  Franzel and van Houten (1992), 6. Heisey and Smale (1995), 7. Corbett et al. (1998), 8. Other geographic information systems (GIS) data, 9. Thorne (1998), 10. Stoorvogel andSmaling (1990), and11. Thorne andCadisch (1998). 3. CP = crude protein, and DM = dry matter. 4. MBW = metabolic bodyweight of livestock. 5. OM = organic matter. 5.7 Potential impacts of the selected interventions 5.7.1 Use of collected weeds of the maize crop for livestock feeding Table 27 summarises the biophysical impacts of weed collection in the smallscale intensive (SSI) and Table 28 summarises those in medium-scale intensive (MSI) systems. Both scenarios are based on a moderate level of weed infestation. In practice, impacts are likely to be highly variable because levels of infestation change from season to season and year to year. Thus, whilst weeding is undoubtedly beneficial for crop performance when infestation levels are moderate to high, it is difficult to see how the practice might form part of a systematic  Franzel and van Houten (1992), 6. Heisey and Smale (1995), 7. Corbett et al. (1998), 8.Other geographic information systems (GIS) data, 9. Thorne (1998), 10. Stoorvogel and Smaling (1990), 11. Thorne and Cadisch (1998), and 12. Jahnke (1982). 3. CP = crude protein, and DM = dry matter. 4. MBW = metabolic bodyweight of livestock. 5. Milk yield is assumed to be zero because farmers are unlikely to target it as a priority intervention in this system. 6. OM = organic matter. feeding strategy. On farm, feeding strategies are generally very flexible, making use of materials when available and substituting these when they cannot be found or purchased. As a result, weeds, when available, may just be used as a more convenient replacement for other collected feeds (such as off-farm/roadside grasses in Tanzania). In this situation, benefits are unlikely to be seen in improved livestock performance.In terms of daily production, effects on milk yields are small. However, over the course of one year an increment in milk of around 70 litres might be realised.  Franzel and van Houten (1992), 6. Heisey and Smale (1995), 7. Corbett et al. (1998), 8. Other geographic information systems (GIS) data, 9. Thorne (1998), 10. Stoorvogel and Smaling (1990), 11. Thorne and Cadisch (1998), and 12. Jahnke (1982). 3. CP = crude protein, and DM = dry matter. 4. MBW = metabolic bodyweight of livestock. 5. Milk yield is assumed to be zero because farmers are unlikely to target it as a priority intervention in this system. 6. OM = organic matter.The availability of thinnings taken from the maize crop between emergence and tasselling has been viewed as an option for producing larger amounts of fodder from the maize crop. Indeed, many farmers in the Kenya highlands and probably elsewhere may plant up to six seeds in each planting hole with the express purpose of removing some of the seedlings later (Lukuyu 2000). In This material is used to supplement the existing, wet season feeding regime. This feed is allocated to a dairy cow. 2. CP = crude protein, and DM = dry matter. 3. In some cases a slight increase in dry season feed availability is predicted because of the sparing effect of using weed biomass during the wet season. MBW = metabolic bodyweight of livestock. 4. Main impact is during the rainy season. This figure is averaged over the year. 5. OM = organic matter. addition to providing fodder, this allows the most vigorous seedlings to be allowed to develop and can also help the farmer to adjust cropping density to water availability. Most indications are that a denser initial planting followed by more intensive thinning during crop development has little effect on overall grain or stover yields (e.g. Shirima 1994, Methu 1998). Other practices relating to the removal of fodder biomass from the maize crop include stripping and topping (see the relevant section of the literature review). These can also provide fodder during the crop's growing season. However, the amounts produced are generally less than those provided from thinnings and, because these techniques involve the removal of material from the plants that will ultimately be harvested, they are also likely to 1. Assumes a moderate level of weed infestation producing 1.5 t DM/ha per annum. This material is used to supplement the existing, wet season feeding regime. This feed is allocated to a dairy cow. 2. CP = crude protein and DM = dry matter. 3. In some cases a slight increase in dry season feed availability is predicted because of the sparing effect of using weed biomass during the wet season. MBW = metabolic bodyweight of livestock. 4. Main impact is during the rainy season. This figure is averaged over the year. 5. OM = organic matter. result in lower yields of dry stover. As this results in the redistribution of feed from the dry season to the generally less critical wet season, the evaluation presented here has been based on thinnings.Table 29 illustrates the predicted biophysical impacts of high-density maize plantings followed by subsequent thinning in small-scale, intensive farming systems and Table 30 in medium-scale, intensive farming systems. Some care should be taken in interpreting these results as it is likely that thinning is already quite widely practised and therefore accounted for, to some extent, in the baseline data. Nevertheless, as the level of aggregation used in generating these predictions is high, it is considered unlikely that this factor would introduce significant extra error into the percentage changes recorded here.Tables 29 and 30 show that thinning does indeed appear to introduce a significant benefit in terms of fodder availability, if not quality, during the rainy season and thereby promotes an increase in milk production. This is achieved without significant negative impacts on grain yield and dry season fodder availability although, as in the case of the use of weeds removed from the maize crop, it has been assumed that farmers do not merely use maize thinnings to replace less readily available feeds. Thinning, from a productivity point of view, appears to offer some promise. In addition it is quite an adaptable technique that can be adapted for different situations. For example, Lukuyu (2000) reports that, during thinning operations, farmers normally leave the two strongest seedlings in each planting hole to develop. However, when fodder is scarce only one seedling may be left in order to produce extra feed. The approach incurs a number of extra costs in terms of seed cost and labour. The implications of these are examined in the economic surplus model below.Despite relatively poor nutritive value, dry maize stover may be the only available basal feed during the drier parts of the year in much of ESA. Its importance, therefore, should not be underestimated and it is likely to form a significant component of dry season feeding strategies in all four systems under study. Methu (1998) suggests that yields of up to 5 t/ha of dry stover dry matter are not uncommon in the Kenya highlands. More importantly, because of the dry nature of the feed it is easily conserved for use in times of fodder shortage, something that does not apply to many of the other feeding interventions in the small-and medium-scale systems. However, because of the poor quality of maize stover, options for improving its contribution to livestock productivity are generally limited to making more effective use of it in the diet. One possibility for improving the extraction of nutrients (by chopping/soaking) is considered as a separate intervention in the next section. Others include supplementation, improved management of feed resources over the dry season as a whole and offering in excess to allow selection of the more nutritious components. This evaluation excludes the latter because its feasibility is restricted to situations where feed shortage can be reasonably guaranteed not to occur. A farmer embarking on a systematic strategy of excess feeding is assuming, at the beginning of the dry season, that the next rains will appear at a predetermined time. Unfortunately, the timing and extent of rains in ESA appears to be increasingly unpredictable (IPCC 2001).Tables 31 to 34 illustrate the biophysical impacts of a range of feeding strategies that aim to make more effective use of maize stover-based diets in dry seasons. Although these are likely to vary in detail by system and country, they all aim to make effective use of the feeds that are likely to be available during the course of a dry season, including the planned use of supplements, particularly for dairy cattle. The bulk of the responses to these interventions may be attributed to the use of supplementary feeding in the more intensive systems. The acceptability of this strategy to the user will depend largely on the balance between the cost of supplementation and returns from milk sales. Nevertheless, even in the intensive systems, a proportion of the benefits is due to better timing of use of the better quality feeds. Kaitho et al. (2001) suggest that the reallocation of concentrate into early lactation can produce as much as 611 litres of extra milk per cow per year. This is a low-or even no-cost option that could also be applicable to the semiintensive and extensive system.Dry maize residues are lignified to a greater extent than green fodder and contain much less protein (typically around 50 g CP/kg of dry matter). Chopping and soaking of stovers are two physical treatments that have been practised with a view to improving the accessibility of the relatively limited amount of substrate material in maize stover to the rumen microflora. Chopping also reduces the ability of cattle to select the more nutritious components of the stover (i.e. leaves and stems) at the expense of the less nutritious leaf sheaths. This can be of benefit where feed availability is restricted. However, in many situations there will be important labour implications for these practices, in addition to dry-season water availability concerns for stover soaking.Table 35 illustrates the likely biophysical impacts of stover chopping and soaking in the medium-scale intensive farming system. These would appear to be restricted to a small increase in productivity related largely to the improved utilisation of the available feed resource (i.e. less wastage of offered feeds). Under normal circumstances, feed refusals obtained from feeding unchopped stover might be included in the manure compost produced from the stall. Impacts on soil fertility of reduced inclusion of stover refusals in manure compost are likely to be very small because of the recalcitrant nature of stover and may in any case be compensated for by the addition of small amounts of leaf litter bedding.Perhaps not surprisingly, it is the replacement of maize with improved fodder crops that offers the potential for major impacts in terms of improved livestock productivity although, of course, this has an equally major negative impact in terms of maize production. Table 36 summarises the biophysical outcomes of adopting this strategy. The scenario assumes that there has been no associated increase in the use of supplements for the dairy cows associated with this strategy. This is perhaps unrealistic because the strategy represents quite a radical degree of specialisation for the farmer. Increasing supplementation by 1 kg/lactating cow per day, in combination with replacement fodder, might be expected to produce between 2 and 4 extra litres of milk each day, depending on the quality of the supplement used.It is likely that considerable economic benefits will be associated with the replacement of maize with forage crops such as napier grass or napier-legume intercrops. However, a cultural resistance is likely from many individuals because this represents the replacement of a staple with a cash crop. Furthermore, adoption of such a strategy would require the household to purchase maize that, with widespread adoption, could lead to imbalances between the supply of and demand for the crop. Therefore, it is suggested that this strategy may be acceptable to farmers in the SSI system who are developing their enterprises through specialisation and particularly to those who have alternative sources of income. The evaluation also suggests some improvements in soil fertility associated with increased manure compost production although no manure management was included in the scenario. It is predicted that these would eventually and partially compensate for the reduced land area allocated to food crops. The move towards a more specialised market-oriented production system is also likely to be accompanied by more systematic applications of manure and inorganic fertilisers. Ultimately, in certain situations in peri-urban systems with well-distributed rainfall, for example, use of manure and compost on vegetable crops rather than fodder crops may be a more financially attractive option.Intercropping is widely practised in the maize crop in East Africa, less so in southern Africa. Even a simple two-species intercrop is a relatively complex ecological system and the outcomes of adopting intercropping systems are not always easily predictable. Moreover, in practice, many of the systems described as intercrops in East Africa are in fact multiple cropping systems of three or more species and even the mixing of varieties within species. For the purposes of an 1. Fifty percent of the area cropped with maize is replaced with napier grass. The forage produced is fed to dairy cattle associated with the holding. 2. CP = crude protein, and DM = dry matter. 3. MBW = metabolic bodyweight of livestock. 4. OM = organic matter. analysis such as this, conducted at a high level of aggregation, it has been necessary to assume that the results of intercropping are:• DM yields per unit area are increased because of more effective exploitation of resources, although the individual yields of the component crops may be reduced in comparison with a monoculture grown under identical conditions. • Some compensation for the increased nutrient demands of an intercrop is provided by the N-fixing abilities of a leguminous companion species and their lower nutrient demands generally.Tables 37 and 38 show the biophysical impacts of intercropping in the medium-scale, semi-intensive and extensive systems. 3. CP = crude protein, and DM = dry matter. 4. MBW = metabolic bodyweight of livestock. 5. OM = organic matter.On-farm management of manure is often rather poor in ESA. It is common to see stalled livestock standing in wet manure with no attempt made to reduce losses of volatile nutrients via leaching. Some farmers do practice removal of voided manure to an area where it can be composted with other organic wastes, and the wider adoption of this technology, together with properly planned applications, could preserve a considerable quantity of nutrients within the system.Tables 39 and 40 show the predicted biophysical impacts of adopting improved manure management strategies in the small-and medium-scale intensive systems.The absolute benefits are not large. However, improved manuring is a relatively long-term strategy that serves to improve a wide range of soil characteristics (such as mechanical properties and water-holding and cation-exchange capacities) with implications for stability of outputs as well as absolute yields. These are not easily accounted for in an exercise conducted at this level of aggregation. Better responses to manure than those implied here have been observed experimentally but may be based on unrealistic application levels (in excess of 3 t/ ha), which are unlikely to be realised through better manure management practices alone.  1. Improved management of voided (i.e. collection and augmentation with other organic material) cattle manure to reduce volatile and leaching losses. 2. Predictions for grain yield reflect the longer term impacts of gradual improvements in soil fertility. 3. CP = crude protein, and DM = dry matter. 4. MBW = metabolic bodyweight of livestock. 5. OM = organic matter.The potential impacts of improvements in maize stover digestibility were briefly examined, at a level that might be achievable through a programme of genetic improvement, on milk production from the small-and medium-scale, intensive systems. Simulations examined the impacts of changes in stover digestibility on the protein and energy supplied to dairy cattle by diets based on maize stover supplemented with dairy meal. Table 41 shows that, assuming a 5% increase in digestibility, the individual benefits of this technology are likely to be small. Although maize stover forms most of the dietary dry matter, its nutrient and energy concentrations are low and, therefore, the impacts on milk production levels of the improved digestibility are unlikely to exceed 50 mL per day. However, scaling up for each study system suggests that even this small increase could deliver a considerable improvement in milk supplied by the systems (see Table 58). There is some uncertainty as to whether these increases would be realisable in practice, because individual farmers may not change to varieties with 'improved' stover quality for such a relatively small payback. It could be assumed, for example, that an increase in individual milk production of around 500 mL per day could feasibly drive uptake of improved genotypes; however, simulations suggest that this would require stover digestibility coefficients of around 93%, a level that is unlikely to be achieved through genetic improvement or indeed any other current technology.The information derived from the modelling studies, summarised in Tables 27 to 41, was scaled up to the system level for the five countries in the following way. To delineate the maize-based, mixed farming system areas in the region, the livestock systems classification of Seré and Steinfeld (1996) as modified and mapped by Thornton et al. (2002) and Kruska et al. (2002) was used. From this, the areas defined as 'mixed rainfed systems' were extracted and overlaid with the areas with a length of growing period (LGP) in excess of 120 days per year to give the total area of land where maize is grown in mixed systems. It was thus assumed that maize could not be grown in areas with less than 120 LGP. These mixed maize areas were overlaid with the human population density layers developed in Reid et al. (2000) and the ILRI cattle databases, to provide the information by country by system shown in Table 42. To estimate the area of maize grown in each system in each country, it was assumed that the ratio of cultivated land to land under maize was constant across systems within a country. To estimate this ratio, the total area of cultivated land for each country was taken from the data set of Wint and Rogers (1998) and the total area of land under maize from FAO (2001). The resulting maize areas by system by country are shown in Table 42. Note that for RSA, estimated maize areas calculated in this way exceed the estimated cultivated areas  Wint and Rodgers (1998). 4. Area = area of the zone, defined as the overlap of 'mixed systems' and length of growing period >120 days (see Kruska et al. 2002). 5. Cultivated area = estimated area of the zone under cultivation (Cultivation % * Area). 6. Estimated maize = maize area by system, calculated as (FAOSTAT country figures for maize in 2000 * Cultivation % * Cultivated area)/Total cultivated area. For RSA, estimated maize areas calculated in this way exceed the estimated cultivated areas in each system, so the maize area was set equal to the cultivated area, which is obviously unrealistic. The reason for this is that most of the maize in RSA is not grown in these four systems (the commercial sector etc.); it may also be that a cut-off of 120 days length of growing period for maize growing in RSA is somewhat high. But without detailed maize distribution data for the country, it is not possible to say whether the cultivated areas in each system are under-or overestimated in the table above.in each system. The reason for this is that most of the maize in RSA is not grown in these four systems, unlike the situation in the other study countries. In addition, a cut-off of 120 days LGP for maize in RSA may be rather high. To ensure consistency, the maize area was set equal to the cultivated area, which is obviously unrealistic but, without detailed maize distribution data for the country, it is difficult to assess whether the cultivated areas in each system are under-or overestimated.These data are mapped in Figures 3 to 6: Figure 3 shows the farming systems in these mixed maize areas, Figure 4 the cattle density, Figure 5 the human population density in 2000, and Figure 6 in 2020. Note that areas are classified as humid/subhumid if LGP is greater than 180 days per year, and as arid/semi-arid if LGP is less than, or equal to, 180 days per year (Seré and Steinfeld 1996). These data were used to scale up the household-level impacts derived in the previous section, and the results are shown in Tables 44 to 58. To assist the reader track what is where, Table 43 lists the tables that describe the baseline data, the predicted household impacts, and the aggregated and predicted impacts by system   and by intervention. Thus the baseline data for the SSI systems are located in Table 23; the household impacts of improved manure management in this system can be found in Table 39 for the countries where this was judged to be viable (in this case, Kenya, Tanzania and Zimbabwe); and the aggregated impacts are shown in Table 48.The final step in this analysis was to value the benefits of these aggregated impacts, in the light of any research and extension costs that may be associated with the interventions and any extra costs or savings that farmers may incur because of them. This valuation was carried out using the economic surplus model (Alston et al. 1995). A partial-equilibrium, comparative static model of a closed economy was used. Assuming a closed economy implies that the adoption of a cost-  1. MSE = medium-scale extensive, MSI = medium-scale intensive, MSSI = medium-scale semi-intensive, and SSI = small-scale intensive. 2. K = Kenya, M = Malawi, S = Republic of South Africa, T = Tanzania, and Z = Zimbabwe.reducing or yield-enhancing technology increases the supply of a commodity. This implies that there is little or no international trade in the commodities concerned, so that an increase in supply reduces both the cost of the commodities to consumers and the price to producers. The analysis here uses the simple case of linear supply and demand curves with parallel shifts. This model is described in considerable detail elsewhere (Alston et al. 1995;Kristjanson et al. 1999) but in sum, adoption of a new technology is assumed to shift the supply curve of the product (such as maize or meat) upwards, resulting in a new equilibrium price and quantity of the product marketed. Gross annual research benefits are calculated as the total increase in economic welfare (change in total surplus) and comprise both the changes in producer and consumer surplus resulting from the shift in supply.Consumers are better off because they consume more at a lower price. Producers may be better off too, because increased supply lowers their per unit costs of production; although they are receiving a lower price for their product, they may be selling more and thus further increasing their benefits.  1. CP = crude protein, DM = dry matter, and OM = organic matter.             To specify fully an economic surplus model, various pieces of information are required, in addition to the aggregated productivity impacts expected:• For each intervention, an estimate of the cost of doing the research and extending the results to farmers; • Changes in gross margins or net returns at the household level arising as the result of adoption of the intervention (e.g. whether it increases input use); • Information on the adoption curve, in terms of the time lag to maximum adoption and the ceiling level of adoption for the intervention; and • Parameters that describe the supply and demand curves for each commodity-the elasticities of supply and demand. 2Monetary values of the impact outputs (Table 13) used in the analysis are shown in Table 59. Feed quality and quantity variables are expressed in terms of meat and milk equivalents and are included in the analysis. The soil fertility variables are not currently accounted for in this analysis. These might be incorporated in a number of ways, most conveniently by valuing them in terms of direct impacts on maize and feed yields in subsequent years. To do this, it would be necessary to run multiple-season simulations of maize and fodder production. As yet, this has not been done, although it would mean that the economic surplus modelling could still be carried out in relation to traded commodities (maize, meat and milk).Table 60 shows the baseline values used for the research and extension costs associated with the eight technologies, and the time to maximum adoption and the maximum adoption rate for each, by system. Figure 7 shows a diagram of the research and extension process, together with the parameters needed for each intervention in each system. Research and extension costs were originally assigned values on a scale from zero to high and then these were assigned monetary values per year per system per country. Of course a great deal of uncertainty is associated with these figures and because of this they are the subject of sensitivity analysis in     and a discount rate of 5% in accordance with the discussion in Alston et al. (1995) on this issue. Table 64 shows the results of selected sensitivity analysis in terms of changes in the present value (PV) of the net benefit stream to 2020 for seven scenarios involving changes in the discount rate, research and extension costs, ceiling adoption rates and on-farm adoption costs. To aid comparison, Table 65 shows the ranks of the present values in Table 64 for these scenarios. Results are discussed briefly by intervention below.In both the SSI and MSI systems, the use of collected weeds of the maize crop for livestock feeding showed reasonable returns. The total benefits arising from incremental milk and maize production (see Tables 44 and 49) are respectable and this is a low-cost intervention: no research needs to be done and the costs arise from extension costs and the costs of adoption, which for this intervention are assumed to be entirely to do with labour. The benefit-cost ratio (BCR) is greater than 6-in other words, the discounted benefits to 2020 are more than six times the discounted costs (Table 63). As might be expected, doubling or halving extension costs had only limited impact on the PV; even tripling the adoption costs per hectare reduced the PV by only 25%. Given the nature of this intervention, results are highly sensitive to adoption rates (columns 7 and 8 in Table 64). That the potential returns to this intervention in the MSI are greater than in the SSI system is due partially to the fact that the total maize area and cattle numbers in the former are somewhat greater than in the latter (Table 42).Economic benefits arise from this intervention because of extra milk production (see Tables 45 and 50). In both MSI and SSI systems, the BCR is greater than unity   1. Systems: SSI = small-scale intensive, MSI = medium-scale intensive, MSSI = mediumscale semi-intensive and MSE = medium-scale extensive. 2. Scenarios: DR = increasing the discount rate from 5 to 10%, R&E*2 = doubling all research and extension costs, R&E/2 = halving all research and extension costs, Adop*2 = doubling the adoption ceiling rates, Adop/2 = halving the adoption ceiling rates, Cost-0 = setting all adoption costs to 0, and Cost*3 = tripling all adoption costs. and the PV of net benefits to 2020 is about US$ 12 million for the SSI system and US$ 14 million for the MSI system (Table 63). Sensitivity analysis shows that the potential returns are quite sensitive to changes in the discount rate used, to the adoption ceilings used and to the on-farm adoption costs (Table 64). Potential returns are much less sensitive to research and extension costs, however, which for this intervention are relatively modest. As for the 'use of weeds' intervention above, potential impacts in the MSI system in Kenya, Tanzania, Malawi and Zimbabwe are somewhat greater than in the SSI system in Kenya, Tanzania and Zimbabwe. 1. Systems: SSI = small-scale intensive, MSI = medium-scale intensive, MSSI = mediumscale semi-intensive and MSE = medium-scale extensive. 2. Scenarios: DR = increasing the discount rate from 5 to 10%, R&E*2 = doubling all research and extension costs, R&E/2 = halving all research and extension costs, Adop*2 = doubling the adoption ceiling rates, Adop/2 = halving the adoption ceiling rates, Cost-0 = setting all adoption costs to 0, and Cost*3 = tripling all adoption costs.For this set of interventions as a whole, across all the systems, the PV of net benefits is an order of magnitude greater than for any other intervention, although there are some logical system differences. For the MSI and SSI systems, the potential benefits arise because of incremental milk and maize production; in the MSSI and MSE systems, the benefits are due to incremental meat production. For the MSI and SSI systems, the BCR is very high, indicating that large potential gains are associated with this intervention (Table 63). Potential returns are highly sensitive to the discount rate used, and to the adoption ceilings used (Table 64).Returns are much less sensitive to the research and extension costs (which are relatively modest) and to the on-farm adoption costs (made up of some extra labour and input costs). This suggests that for these systems, potential returns could be strongly associated with effective extension activities, provided that the on-farm adoption costs of these technologies do not pose a serious constraint to uptake. The situation for the MSSI and MSE systems is somewhat different. Potential returns in the MSE system are not large enough to cover adoption and research and extension costs. In the MSSI systems, net benefits are small, as are the costs, but highly sensitive to adoption rates in particular.This intervention, assessed in relation to the MSI systems, provides benefits through small increases in milk production. These benefits do not outweigh the costs involved, however; extension costs are relatively large, although the research costs are modest (Table 63). Interestingly, the results of the sensitivity analysis indicates that potential returns are not very sensitive to adoption rates but highly sensitive to on-farm adoption costs. Whatever the costs involved, the potential biophysical benefits do not appear to be large enough to outweigh even the modest investments in research and extension required for this intervention to be promoted and adopted.The results of this intervention in SSI systems are of some interest. Table 63 shows clearly that the potential returns are highly negative to 2020. The reasons can be found in Table 47; benefits of this intervention arise from increased milk production but these are balanced by negative impacts on country maize yields. At current prices for maize and milk in Kenya, Tanzania and Zimbabwe, the value of the increased milk is simply outweighed by the costs of the lost maize. In fact, the price of milk relative to maize would need to increase by a factor of 6 for the BCR to exceed unity for this intervention. Sensitivity analysis indicates that large changes in adoption costs and research and extension costs have relatively little impact on potential returns; the returns are much more sensitive to adoption ceilings. But in any case, the results suggest that replacing maize with fodder crops is unlikely to show positive returns at the farm level under current pricing regimes.Intercropping in the MSSI and MSE systems produces benefits through small increments in meat offtake, decreases in monocrop maize yield increments and larger increases in intercrop increments. There are also moderate impacts on soil fertility in terms of C and N conserved each year (Tables 55 and 57). Intercropping produces substantial net benefit streams, with a BCR of 5 to 6 in both systems. Overall, the potential impacts of this intervention are the second highest of all the interventions looked at, after improved feeding systems (Tables 63 and 64). The results are also relatively robust. The PVs are not very sensitive to research and extension costs (which are modest for this intervention), or to the on-farm adoption costs (so added costs are not likely to be a great constraint to the uptake of intercropping). The PVs are much more sensitive to changes in the discount rate and, particularly, the adoption ceilings used. Doubling the adoption ceiling, from 5 to 10% of the appropriate domains, nearly doubles the resultant PVs for these systems.Improved manure management in the SSI and MSI systems provides benefits through maize yield increments and through soil fertility impacts, which may be large. The gross benefits of this intervention are substantial (Table 63) but are balanced largely by the costs involved. These are made up of medium research costs, high extension costs and reasonably high on-farm adoption costs. Even so, the PV of the net benefit stream is positive, and the BCR exceeds unity for the MSI system. The situation for the SSI system is more equivocal: the PV is only just positive and the BCR barely exceeds unity. The results are highly sensitive to discount rate, partially because a reasonably long time to ceiling adoption levels is assumed for this intervention. The results are also highly sensitive to research and extension costs, to on-farm adoption costs and to ceiling adoption levels in SSI and MSI systems. Of all the interventions assessed, this is the one that appears to be most sensitive to the assumptions used in the analysis and, as such, is deserving of further study. More detailed assessment of research and extension costs and of onfarm adoption costs would definitely help in defining potential benefits more precisely. In addition, as the longer-term soil fertility impacts are not valued in this analysis, the returns to this intervention are likely to be underestimated; there may be increasing increments of maize yield over time with such management. As noted above, multiple-season simulations of maize productivity would capture these added benefits and these simulations should be done in the future to finetune these analyses.This intervention was assessed in the SSI and MSI systems (Table 58). Benefits accrue through moderate milk yield increments in these systems. As can be seen (Table 63), these benefits are very modest and, when weighed against the costs involved, made up mostly of substantial research and extension costs incurred between now and 2020, the PV of the benefit streams are negative and the BCRs are less than unity for both systems. Sensitivity analysis suggests that results are highly sensitive to research costs but much less so to adoption ceilings and onfarm adoption costs (which are low). Even with halving all research and extension costs, the PV of net benefits is still heavily negative. These results suggest that, technically, the case for pursuing this intervention targeted at the maize-based mixed systems is less than compelling. The impacts on livestock production of improved feeding quality of maize stover would appear to be too small at the household level to boost adoption of improved varieties of maize in general.The results of the economic surplus modelling contain some interesting indications as to where potential impacts of research and extension on maize in the mixed systems of the five countries might be appropriately targeted. Improved feeding systems appear to offer substantial potential for smallholders, particularly those in the more intensive systems. Significant research and extension costs are associated with these but the potential benefits far outweigh these costs. Promoting the use of intercropping in the more extensive systems where this is not already practised also offers substantial net benefits. The extension effort involved may be considerable but again the potential impacts at the household level are such as to outweigh these costs. Use of weeds and improved green stover management have modest research and extension costs but could provide modest net benefits in the more intensive systems. The case of improved manure management is less clear cut. In the MSI systems, there may be substantial benefits to be reaped but less so in the SSI systems, although the results are highly sensitive to the assumptions made. This suggests that research and extension on this intervention would need to be very well targeted for appropriate societal benefits to accrue. Little evidence could be found of net benefits for treating dry maize stover (high research and extension costs coupled with limited benefit at the farm level), or breeding/ selection for improved stover digestibility. Current price regimes are not conducive to replacing maize with fodder crops in the SSI systems and are not likely to become so in the future. These results could be scaled up to the whole East and southern Africa region. This study was concerned with case studies in just five countries, for logistical and resource reasons. There is a considerable additional area in the region that is similar, in terms of maize-based mixed systems. These areas are mapped in Figures 8-10, with respect to the farming systems and human population density in 2000 and to 2020. Some salient characteristics of the case study countries compared with the entire region are presented in Table 66. The case study countries account for some 32% of the land area in maize-based mixed systems in East and southern Africa as a whole. They also account for 68% of the maize grown, 40% of the population and 48% of the cattle.For research and extension programmes with a limited budget, there are viable options for all system types. As might be expected, potential benefits for the SSI and MSI systems are much greater and there are more options for the more intensive systems, and activities could focus on a variety of interventions (Table 67). Viable options for the MSSI and MSE systems are much more limited and in general involve much greater extension than research effort. In fact, three features of Table 67 are particularly striking:• The number of viable options decreases as system intensity decreases • The amount of research effort that is needed decreases as system intensity decreases and • The amount of extension effort needed for many of the options is consistently high, whichever systems are being targeted.      Although these may come as no surprise, the importance of extension is highly relevant and topical. Extension has been seen for far too long as the 'poor relation' of 'proper' research but these results suggest that substantial and sustained and targeted extension efforts are required if potential benefits are to be realised. The recent history of agricultural research in sub-Saharan Africa bears out this message very clearly. On the research side, the only feasible option that appears to require substantial research investment is improved manure management and this would apparently need to be quite narrowly targeted to generate substantial benefits to society as a whole.In addition to the usual uncertainties that abound in analyses such as those presented above, an important omission in the above is that of the soil fertility impacts. Inspection of Tables 44 to 58 indicates that soil fertility impacts arise for the following interventions: improved feeding systems, replacement fodder, intercropping and manure management. Leaving out the replacement fodder intervention, which is not economically feasible, and given 100% adoption, these tables indicate that improved feeding systems in the SSI and MSI systems could conserve about 70 thousand tonnes of C per year and 13 thousand tonnes of N per year; intercropping, 20 thousand tonnes of C and 5000 t of N per year; and improved manure management, 146 thousand tonnes of C and 9200 t of N per year. Even with realistic adoption rates, which might decrease these values by a factor of 20 or more, these are still significant quantities and may have substantial additional incremental impacts on fodder and maize yields in subsequent years. Unfortunately, the simulation of what these impacts may be over 10 years or so is not a trivial matter but it is hoped to be able to do these in the future. A further caveat should be expressed, although it is an obvious one. The economic viability of the various interventions is just one criterion that might be used in assessing potential impacts of research and extension interventions for resource allocation decisions. There are, of course, many others that could and can be used, including criteria relating specifically to poverty alleviation, environmental impacts and human capacity building, to name but a few (see Randolph et al. 2001 for a summary discussion of some of these).Despite various limitations, this analysis has provided some insights into the nature of the interventions that could assist in improving the lot of smallholders in the region who depend on maize-based, crop-livestock systems. Much could be done to improve some of the analysis, particularly the longer-term impacts of some of the interventions. Nevertheless, we now have the tools to start to assess different interventions at the system level. The analysis reported here has provided information that can help to prioritise research and extension activities that should be able to contribute to widely held development goals.Table A3. Aggregation of benefits: All commodities, all countries, in the case of improved feeding systems in small-scale, intensive systems.Total benefits Total benefits Total benefits Total benefits Total benefits Total costs Net benefits Discounted fodder grain meat milk (US$ x 10 "}
\ No newline at end of file