Toward Sustainability: A Plan for Collaborative Research on Agriculture and Natural Resource Management (1991)

The proposed Sustainable Agriculture and Resource Management (SANREM) program is distinguished from its Collaborative Research Support Program (CRSP) predecessors by its focus on the sustainability of agroecological systems. Previously established CRSPs, and international agricultural research efforts in general, have focused on the development of technologies to increase the production of particular commodities. The commodity focus has enabled researchers to build interdisciplinary teams and methodologies, to strengthen institutional structures around the world, and to bring local experience and needs to the attention of the global research community. Not incidentally, commodity-centered programs have also yielded important insights into a variety of agronomic considerations—germplasm conservation, nitrogen fixation, rotational effects, and pest population dynamics, to cite only a few examples —that are critical to sustainable agriculture in their particular agroecosystems. The commodity focus has also enabled researchers to begin to define the social and economic issues associated with those particular commodities and the regions in which they are grown. The SANREM program must incorporate and build on the substantial record of achievement that those research programs have compiled.

The emergence of sustainability as an organizing concept and as a research objective is in itself evidence that, however effective in its special

This appendix is based on conclusions and research priorities identified by participants at the National Research Council's Forum on Sustainable Agriculture and Natural Resource Management, held in Washington, D.C., on November 13–16, 1990. Working groups discussed priorities in each of the four agroecosystems described below, and a fifth group addressed priorities across agroecosystems.

applications, commodity-oriented research is limited in its ability to embrace all the factors that influence the long-term health, productivity, and stability of the agroecological system as a whole. In focusing scientific attention on the productivity of a given plant crop, even broadly conceived research tends to neglect other crops, the full range of environmental influences on the crop, the environmental impact of the crop, the social and economic causes and effects of cropping systems, and the role of the crop in achieving a balanced and equitable system of land use. These factors also act as feedback mechanisms; they can eventually affect, both positively and negatively, the crops under scrutiny. The systems approach works to harness this understanding for the long-term well-being of the crop, people, and the system as a whole.

Agroecosystem research recognizes that the delineation of system boundaries implies a certain degree of flexibility. Boundaries must be defined clearly enough to allow for rigorous study, and loosely enough to take into account less immediate but still relevant factors affecting the system. In this sense, boundaries can vary depending on the scope of the hypothesis being tested. An investigation of the effects of intercropping on soil microbial activity in a single field, for instance, demands a different, though no less legitimate, scale of research than does a landscape-level investigation of the effects of cropping patterns on water quality. The agroecosystem approach enables researchers not only to adjust their focus to the scale appropriate for the hypothesis at issue, but to integrate hypotheses so that they may illuminate one another and the working of the system as a whole. The agroecosystem, in short, will be a fundamental concept in SANREM research, not only as an object of study, but as a way to study. It will serve as a tool to organize ideas, hypotheses, methods, and results, and ultimately to gain insight into principles of sustainability.

The resource base and the human population that relies on it are at greatest risk in several primary global agroecosystems: the humid tropics, semiarid range and savannah, hill lands, and input-intensive agroecosystems. In establishing and building the SANREM program, researchers will invariably focus on these systems. They offer the greatest potential for results that (a) are critical to environmental well-being, (b) can be broadly applied, (c) are relevant to great numbers of farmers, and (d) can help to define more precisely the characteristics of sustainability.

The above classification of primary global agroecosystems is not homogeneous. The first two systems are defined primarily by climatic and vegetational factors, the third by overriding topographical characteristics (slope, aspect, and elevation), and the last by input levels. Broad as they are, these categories necessarily obscure the incalculable diversity of local conditions within each—their ecological circumstances, characteristic biological diversity, historical land-use patterns, and cultural contexts. Moreover, certain

sites legitimately belong to more than one category. Other systems, less extensive but nonetheless important in terms of human welfare, biodiversity, and other aspects of sustainability, may fit none of these categories. Such unavoidable shortcomings aside, this classification allows the broadest and most effective identification of problems, possibilities, gaps, and commonalities in the complex undertaking of research on sustainable agriculture and natural resource management.

Humid tropic agroecosystems are located in tropical regions where there is no more than a 3-month dry season and temperature is not a limiting factor for plant growth. The native vegetation in these areas is tropical rain forest. Rain forests once covered some 1.6 billion hectares (ha), principally in the low latitudes of Central America, South America, Africa, and Southeast Asia; smaller expanses existed along the eastern coasts of Madagascar, South America, and Australia. In recent decades, all of these areas have undergone rapid conversion. Approximately half of the former rain forest has been cleared for timber, fuelwood, farming, plantation agriculture, and cattle ranching. The remaining rain forests are concentrated in three large swaths: the Amazon basin in South America, the Congo basin in west central Africa, and the islands of the Malay Archipelago between Southeast Asia and Australia.

Indigenous agricultural techniques have evolved to fit the demands of the rain forest environment. Shifting cultivation enabled small populations of dispersed farmers to raise crops by clearing small plots in the forest, burning the slash for nutrients, raising a series of diverse crops in succession over a period of several years, and then allowing the plot to lie fallow to regain its forest cover and its vegetation-captured fertility. As population pressure has increased, however, shifting cultivation has become more prevalent and its time sequence more compressed. Under such conditions, clearing becomes more frequent and fallow periods decrease or disappear altogether. Large-scale logging, mining, plantation agriculture, and livestock ranching bring more settlers and more intensified land uses to the rain forests. These trends have placed ever greater pressures on the relatively poor tropical soils. Once cleared, the soils are easily eroded, their residual nutrients leached, and their role in the hydrological cycle disrupted. Soil acidity and nutrient deficiency are common chemical constraints to crop production in the humid tropics under any circumstances, and addressing those problems is key to the development of sustainable methods that can ease the pressure on the remaining rain forests, restore already degraded forestland through a variety of agroforestry strategies, and allow for more efficient and intensive cropping of developed land.

In any discussion of the future of the humid tropics, sustainable agriculture must be linked to the causes and consequences of deforestation. The forests of the humid tropics are currently being cleared at a rate that exceeds 10,000 square kilometers (km²) per year. The rate of conversion has roughly doubled since the 1970s, when alarms about deforestation were first widely sounded. The destruction of tropical forests is of concern for three main reasons. First, the destruction of tropical forests currently releases between 25 and 30 percent of annual atmospheric carbon monoxide, carbon dioxide, nitrous oxide, and methane and less dramatic but still important quantities of other greenhouse gases. Although the industrial nations bear most of the responsibility for aggregate atmospheric carbon additions, the destruction of forests is particularly significant because important carbon sinks—tropical forest biomass and soils—are now becoming carbon sources. The models of greenhouse dynamics remain contentious, but they suggest that some of the areas of greatest agricultural production in the United States, especially California, Florida, and parts of the Midwest, are vulnerable to climate modifications. Deforestation also influences carbon emissions through its effects on local microclimates. As surface temperatures increase after conversion (often by more that 10 degrees centigrade), the breakdown of soil organic matter doubles. The consequent release of carbon from soils is orders of magnitude greater than that released by biomass burning. The long-term global changes and the potential dislocations they imply should make the control of deforestation an urgent priority.

The loss of biodiversity is a second consequence of current shifts in land use in the humid tropics. The rain forests, which now cover 7 percent of the earth's surface, are believed to contain at least 50 percent, and perhaps as much as 75 percent, of the total species diversity on earth. Deforestation is consequently bringing about the greatest destruction of the earth's organisms since the Cretaceous extinctions. There are strong ethical and economic reasons to avoid this annihilation of species. The potential economic returns from medicines, latexes, resins, and fibers are important, but they are eclipsed by the importance of wild genera of primary food and industrial crops. The loss of diversity of domesticated varieties has also accelerated as local farmers move off the land or adopt new varieties. The loss of biodiversity increases the vulnerability of both industrial and subsistence agriculture, and it narrows the base for commercial and subsistence plant breeders in developed and developing countries. The structure of U.S. agriculture will ultimately be affected, not only in areas that produce tropical crops (Hawaii, Puerto Rico, Florida, and California), but in virtually all regions in which major food and industrial crops are grown. The diminishment of floral and faunal diversity is cause enough for concern, but the displacement of forest peoples implies also an incalculable depletion of cultural diversity and indigenous knowledge.

Finally, the degradation of soil resources—through erosion, destruction of soil fertility, and loss of lands through urban and industrial encroachment—is advanced in the humid tropical areas. Although it is not the only factor leading to land abandonment and the poor performance of short-cycle tropical agriculture, soil degradation is a major contributing factor.

The above concerns make the careful management of tropical vegetation, soils, and water an urgent priority. Many current production systems, ranging from some kinds of shifting cultivation to industrial agriculture, are unstable under current social and economic conditions. The development of ecologically, socially, and economically viable forms of land use in the humid tropics will require a strategy that builds on their characteristic diversity of the humid tropics and mimics their complex ecological processes.

Research on the humid tropics is expanding as attention is drawn to their status, their role in the global environmental system, and the fate of the people who depend on them for their livelihood. Studies of sustainable agriculture must lead the way in establishing sound principles for land use and conservation in these regions. The following areas and subjects of research are suggested.

• State-of-the-art inventory, classification, and analysis of local/indigenous systems, successful experimental systems, and case studies pertaining to land-resource management. This review would serve as the foundation for, and provide insight into, elaboration of additional elements of a research agenda.

• The possibilities of restoring degraded lands and elaboration of criteria for determining when and what to restore; the possible limits to restoration; the extent to which knowledge and modern techniques are sufficient, individually or together, to restore damaged lands to functioning forests, grasslands, or farmlands. Analysis of the economic costs of land degradation and restoration should be included.

• The development and promotion of general principles and components of land management that sustain land resources under the constraints of tropical ecosystems. This research should involve careful analysis of processes (for example, the management of nutrient cycles and the manipulation of succession) that underlie the sustainability of successful systems and the identification and elaboration of new crops and innovative land-use systems that can help to overcome the short-cycle crop biases that can perpetuate degradation.

• The social forces that drive resource degradation—issues of political economy, accounting of forest goods and services, and policy. Institutional structures that mediate resource use and tenure issues should be analyzed to discover which of them promote careful resource husbandry, and under what conditions they do not. Full account must be made of the value of

forest goods and services, including nontimber forest products, ecosystem services, conservation values, and costs of recuperation.

• Nutrient-cycling patterns and determination of the mass balance of nutrients and water across the full range of humid tropic agroecosystems.

• Issues of sediment additions, water quality, water availability, and water resource management. Given that tropical areas cycle more than 30 percent of the world's freshwater, encompass the largest zones of riparian vegetation, and supply many of the great fisheries of the world, these issues are particularly crucial in the tropics.

• Training U.S. and local scientists. As in other agroecosystems, scientific training is fundamental to research in the humid tropics. Any long-term strategy for improving the productive and protective capacities of tropical environments must develop the local research capacity and strengthen the local institutional support—and these need not be state or official institutions—for careful management of land and water resources. Cooperation among farmers, nongovernmental organizations, and researchers must be a key factor in elaborating new strategies and in providing extension and oversight.

Semiarid savannahs and rangelands are characterized by relatively low annual rainfall. Native vegetation—grasses and grass-like plants, shrubs, and drought-resistant trees—evolved within the limits imposed by the protracted dry seasons typical of these regions. In addition to water availability, soil acidity and inherently low soil nutrient levels act as major constraints on intensive crop production. Irrigation, where technically and economically feasible, can make semiarid lands richly productive, but careful management is required to avoid the long-term problems of salinization, waterlogging, aquifer depletion, surface water pollution, and disruption of hydrological systems.

Livestock grazing is an important economic activity in populated semiarid regions. Grazing by wild and domesticated herbivores is essential to the health of rangeland ecosystems, and traditional pastoral cultures were able to maintain human and ruminant population numbers within, and fit grazing patterns to, the carrying capacity of these lands. Human population growth, however, has placed increasing pressure on many semiarid lands. Overgrazing has increased in frequency and extent, and in some areas it has triggered the positive feedbacks that lead to environmental degradation in grassland ecosystems: decreased vegetative cover, invasion by unpalatable species, declining livestock quality, excessive wind and water erosion, soil erosion and degradation, increased susceptibility to the effects of drought, and ultimately desertification. Sustainability in semiarid regions will depend first and foremost on the recognition of the inherent fragility of semi-

arid lands and the tight relationships among available moisture, soil structure, soil nutrient levels, cropping and livestock patterns, the potential impacts of interventions, and human population pressures.

Semiarid range and savannah is widely, though unevenly, distributed around the world; it covers approximately 50 percent of the earth 's surface. Over half of this area, however, is too cold, too dry, or too distant to support permanent concentrations of humans and their livestock and associated crops. Of the occupied rangelands, sub-Saharan Africa, from the West African Sahel through Sudan and Ethiopia to Somalia, faces the most urgent agricultural and environmental difficulties, and it will be the most important testing ground in the near future for the development of sustainable systems. Although semiarid sub-Saharan agricultural systems may have been sustainable under low-intensity exploitation, demographic, climatic, ecological, and institutional factors constitute threats to sustainability for both the short and long term.

Farmers in semiarid Africa typically grow a variety of drought-tolerant staple food crops in fields around their villages, principally sorghum and millet in pure stands or intercropped with cowpeas. Cash crops include cotton and groundnuts. Generally, households also cultivate small plots around the homestead, where they plant maize and vegetables for home use and market sale (often supplementing rainfall with water from recently drilled village tube wells). These agricultural systems are highly integrated and are frequently maintained with minimal or no external inputs. Farmers plant local crop varieties, rely on bush-fallow rotations where possible, use animal manure to maintain fertility, and use family labor to meet the highly seasonal demand for agricultural labor. Cereal yields are low, and any available marketable surpluses are primarily the result of interyear variations in rainfall. Donkeys, cattle, and small ruminants, sustained on crop residue and rangeland, are a source of manure, and in some cases of income, draft power, and food. Households typically give priority to the production of sufficient staple crops to meet the family' s food needs and allocate remaining, often productive, land to cash crops and livestock.

Under any circumstances, agriculture in sub-Saharan Africa is an inherently complex, interactive, and high-risk endeavor. Sustainable approaches are badly needed because population growth has put serious pressure on the fragile natural resource base. Agricultural development in sub-Saharan Africa, as in most semiarid regions, is constrained by readily identifiable factors: water availability; soil nutrient availability, erosion, physical properties, and organic matter; the institutional and human resource base; and the policies necessary to manage soil and water resources. Where overgrazing of rangelands threatens to induce the disintegration of plant communities and soil erosion, farmers have limited resources to make capital investments, and short-term returns are necessary to make investments attractive.

Sustainability under these circumstances will depend on the capacity of the low-income, medium population density countries of sub-Saharan Africa to improve the institutional climate and develop the soil-water-crop-animal systems necessary for agricultural development. The SANREM program can make substantial contributions to this effort. The observations and recommendations that follow are based on the assumption that the sub-Saharan region offers the best returns on investments in terms of widely applicable principles and effective strategies for sustainable agriculture and resource management in semiarid regions.

Soil and water are central to sustainable agriculture across the semiarid tropics. Priority must thus be given to the development of an integrative and environmentally sound systems approach to soil, water, crop, and animal husbandry in an agroecosystem context that will sustain the natural resource base, with full consideration given to institutional and policy factors (that is, land tenure). The approach developed should not only use current knowledge in an integrative manner, but also develop new knowledge and the means for its dissemination and application. Research procedure should involve the creation of an innovative model to address this overriding priority.

In locating CRSP research projects in the semiarid zone, priority should go to those sites that can demonstrate host country institutional commitment and capacity, broad applicability of potential results to other semiarid regions, and in-country mission involvement (but not extensive mission management support; for example, the Tropical Soil Management CRSP model). Research should be conducted on a watershed level at a minimum.

Proposals for research in this agroecosystem should demonstrate the following capabilities:

• innovative approaches to system modeling that are realistic, workable, and applicable;

• integrated research experience, previous commitment to work in the semiarid zone, commitment of university cost-sharing resources, continuity of staffing, and experience in systems research and management;

• an agroecological research framework that gives full attention to biotic, abiotic, and socioeconomic factors, including analysis of indigenous natural resource management; and

• complementarity and interaction with nongovernmental and private voluntary organizations, other CRSPs, and other international agricultural research organizations.

Mountain agroecosystems constitute approximately 25 percent of the total land surface of the earth, and they contain at least 10 percent of the total

population. Major mountain agroecosystems are found in the Andes of South America, throughout Central America, in the Rockies of western North America, the islands of the Caribbean and Southeast Asia, the Hindu Kush-Himalaya region of South Asia, and the mountains of East and Central Africa. In virtually all of these regions, mountains exist as large humid “islands” in an otherwise arid-to-semiarid landscape, and they serve as a source to major river systems. Although populations in the mountains are relatively low, those of the “highland-lowland interactive system” are high, and they may constitute nearly half of the total population of the earth.

It is difficult to generalize about the mountain agroecosystem, because it incorporates elements of all other ecosystems—from the humid tropic ecosystems on the eastern slopes of the Andes of South America to the arid and semiarid ecosystems of the western Himalaya in South Asia. Above all, the mountain agroecosystem must be viewed as a composite of ecosystems: a three-dimensional environmental mosaic defined by factors of altitude, slope, and aspect, and characterized by agricultural problems encountered across the full spectrum of agroecosystems. In contrast to the relative spatial uniformity of many lowland systems in which traditional agriculture has evolved, the mountain system is defined by a complex terrain that limits the availability of land suitable for agriculture, that underscores the isolation of the farmer, and that highlights the importance of terrestrial-meteorological interactions in providing the water and energy necessary for sustainable plant and animal production.

Mountain agroecosystems and adjacent lowlands are dynamically linked. Water and sediment flow from highland watersheds to lowland river basins. Human population pressures in the lowlands often force more intensive development and exploitation of upland soils, forests, and grazing lands. Although the full implications of these linkages remain subject to debate, it is clear that, at least in some cases, sustainable development of the mountain system may contribute to an increase in the nonsustainability of an adjacent lowland system —and vice versa.

As is the case with many agroecosystems, those of mountains and highlands are poorly understood. Similarly, the potential of mountain systems for sustainable use has not been determined and establishing that potential remains the fundamental challenge. This in turn requires careful assessment of the system's capacity to remain stable in response to external interventions, as well as variable natural processes from within. In the mountains, as in any agroecosystem, this assessment must be based on a thorough understanding of the complex interactions of biophysical and socioeconomic factors. Soil and water, people, institutions and cultures, and economic returns on investments of labor and capital must all be considered in the formulation of appropriate management strategies.

The development of such strategies in hill lands must be based on a dynamic model of mountain agroecosystems that can identify and evaluate alternative strategies prior to their implementation. Priority should therefore be given to the development and testing of such a model, or models. This effort could begin by making use of existing biophysical and socioeconomic concepts and data bases, which would help both to build and evaluate the model, as well as to define more clearly the gaps in existing knowledge. The mountain agroecosystem model should rely heavily on emerging computer-driven information storage, remote sensing, and data analysis technologies. Development and testing of the model should begin with the most fundamental, and potentially unstable, characteristics of the mountain agroecosystem —the soil and water “life support” resource base—and should eventually incorporate all factors, including the socioeconomic and cultural.

Once a reliable model is available, researchers can develop new techniques to evaluate factors relevant to sustainability in mountain agroecosystems, including the suitability of landscape, ecosystem, or socioeconomic units for various management options; mitigation and control methodologies; activity options and alternative agricultural technologies; comparative advantages, in biophysical and socioeconomic terms, of available methods; and the economic, production, and environmental impacts of potential interventions. Only after this phase is completed should actual site-specific development and testing of more specialized models that reflect the great diversity of mountain agroecosystems be undertaken. The ultimate objective is to develop a systems approach to the planning and management of mountain agroecosystems that farmers, resource managers, and institutions can use.

In addition to U.S. universities, this activity should involve established international institutions with demonstrated capability in research on mountain agroecosystems, such as the International Centre for Integrated Mountain Development in Nepal and the University of the Andes in Venezuela. The network of mountain scientists represented by the International Mountain Society should also be used to the extent possible.

In sum, research on mountain agroecosystems should proceed in the following manner:

• Develop and test a dynamic model of mountain agroecosystems.

• Based on that model, develop and test methodological approaches to sustainable development of mountain agroecosystems.

• Prepare training materials and opportunities, including workshops, seminars and short courses, that acquaint planners, managers, and farmers with the potentials and constraints of the mountain agroecosystem and that provide for regular local input into the development and application of the model.

Input-intensive cropping and livestock systems are found around the world. Such systems are characterized by the application of fertilizer to maintain or build soil nutrient levels each year or each crop rotational cycle and by the use of pesticides or biocontrol methods to reduce pest losses to or below threshold levels. Input-intensive systems currently account for the lion's share of world food production. They are found mainly in lowland areas and are dominated by rice, wheat, sorghum, and corn production, particularly in countries facing heavy population pressures. The sustainability of production in such systems is a vital food security and environmental concern.

Input-intensive systems are growing in importance in temperate upland regions and in the savannahs of Africa and Latin America. In the highlands of Central America and on many islands, nearly all food is produced on sloping land, often through very intensive systems that are sustainable only if soil erosion is controlled by producing high levels of crop residues and land cover year-round. Outside irrigated regions, the rain-fed agroecosystems face a distinct mix of technical, economic, and environmental problems.

Because input-intensive systems must contribute much more prominently to total food production if world food needs are to be met, they clearly warrant increased emphasis. The SANREM program should entertain proposals from all geographic regions where input-intensive systems, as defined above, play an important role in meeting regional food needs. The top priority for research on input-intensive systems should be to assess the interactions and implications of efforts to attain higher average yields, especially as they affect long-term productivity of soil and water resources and environmental quality, both on-farm and within the region. To this end, the relationship between attainable yield goals and yield instability may be of great importance from the perspective of food security and, hence, warrant special focus in research proposals. Investigators should also be required to explain how proposed research projects will improve understanding of the roots of yield instability within the region for the crops under investigation, and of the factors that could increase sustainable yield goals.

The proposals should also emphasize the relevance of the proposed research in identifying cropping and animal system technologies that can contribute to higher average yields and improved farm income, without inordinately increasing risks or per unit production costs. Another essential component of the research proposals should be a description of any changes needed in policy, institutions, and infrastructure investments to create and sustain economic incentives and markets.

Because enhancement of the inherent capacity of soil to sustain plant growth is of critical importance in achieving sustainability, investigators should also describe how the proposed research will contribute to the design

of profitable farming systems that are able, over several years, to improve (a) soil physical properties, (b) soil nutrient levels and nutrient-cycling capacity and efficiency, and (c) the ability of the soil to take in and hold available moisture without causing salinity, waterlogging, or adverse effects on off-farm water quality. Equally important, researchers should explain how the proposed research will clarify the impact of agronomic and pest control practices on below-ground soil microorganisms, the levels and virulence of plant and root pathogens, and the significance of soil fauna in nutrient cycling and water retention. Researchers should also describe how they will take into account the spatial variability in landscapes, institutions, and marketing opportunities in the design of cropping and livestock systems.

In meeting these general criteria, each proposal should include the following components:

• a description of the distinct area and agroecosystem in which the research will be conducted and the collaborative efforts that will be undertaken with local organizations and institutions;

• an explanation of the local, regional, and (if appropriate) global significance of the type of cropping systems chosen for analysis;

• a discussion, with a high degree of specificity, of the biological, ecological, physical, social, and economic conditions necessary for sustainability that the proposed research will help elucidate; and

• an evaluation of the importance of socioeconomic, infrastructure, land tenure, and policy considerations in the evolution of cropping practices that may prove unsustainable, and in the adoption of improved production methods that would evolve from successful completion of the proposed research project.

Sustainability in its broadest sense will require the development of management systems that can meet changing human needs in a manner that conserves natural resources and preserves environmental integrity, especially in the various agroecological zones described above. To aid progress toward this end, the SANREM research agenda will and must vary to fit the geographical, ecological, historical, and cultural realities unique to each locale. Progress in all systems, however, benefits from the recognition that they share certain features and that comprehensive scientific understanding requires an appreciation of the similarities across agroecosystems, as well as the differences among them.

Common elements can be identified in the agroecosystems described above, and in other systems that may not fit those categories. These include physical and biological factors—nutrient cycling, biodiversity, soil and water management practices, and disturbance regimes—and socioeconomic

factors—land tenure and property rights, resource policy, infrastructure, gender roles, and economic constraints. Common qualities—in particular, productivity, stability, resilience, and equity—are closely associated with health in each agroecosystem. These commonalities have direct implications for the conduct of research under the SANREM program. The new program must identify ways to focus and promote ongoing research on sustainability issues in other CRSPs; to foster the interaction of indigenous knowledge and scientific methodologies; to further the necessary integration of the disciplines involved in land use and management; and to make local participation a central element of the research process. Perhaps most important, the SANREM program must identify, select, and implement projects that can fill the gaps in current knowledge. Some of these gaps have been identified above, but others will emerge. The search for hidden factors in the sustainability formula will be an important aspect of the SANREM program, and of the systems approach it adopts.

The concentration in this discussion on terrestrial systems ought not to obscure the significance of aquatic systems, in their own right or as they relate to agricultural practices and other aspects of natural resource management. Water itself is a factor common to all ecosystems. A comprehensive scientific approach to the environment in which agriculture is practiced must account for the water resources used in, and the aquatic systems that are affected by, agriculture. As fish play an increasingly important role in the human diet (particularly in developing countries, where they often account for over 40 percent of animal protein consumed), coastal-zone harvesting and aquacultural activities must necessarily be incorporated into the sustainability research agenda. Fisheries and aquaculture entail special considerations, but they are subject to the same principles that govern the sustainability of land-based agriculture; in many regions the two are tightly coupled. Aquatic and agricultural ecosystems are also directly linked by biological and physical processes (the most broadly significant being the cycling of nutrients through waste conversion and feed and fertilizer production); by environmental concerns (especially water quality issues involving soil erosion, siltation, and the runoff of pollutants, fertilizers, and pesticides); and by the prospect of global climate change and its attendant impact on sea levels and biodiversity. These and other considerations point to the need to weigh fully the aquatic component, and its potential contribution, in research designs.

In certain agroecosystems, aquaculture may come to play a direct and highly significant role. Not all agriculture can incorporate aquaculture, but a significant proportion can, and the result can be a more sustainable production system. The entire SANREM program does not need an aquaculture component, but at least some sustainable agriculture systems should be developed with aquaculture as a functioning component of the system. The

existing aquaculture CRSP can provide an excellent data base and a cadre of trained professionals able to bring their experience to the scientific exploration and development of integrated agriculture-aquaculture systems.

The control of pests is another universal feature, common to all agroecosystems, unique in its local needs, and central to the SANREM research program. In its method of investigating and responding to complex environmental phenomena in an agricultural context, integrated pest management provides a model for systems-based research and is itself a vital component of sustainable agriculture.

Underlying this identification of features common to all agroecosystems is the question the effort is meant to address: how can science best serve to inform the issues that sustainability raises? Sustainability is itself a relatively new term, and researchers have only begun to define the structure of the science that describes it. At this point, one can say that there are fairly well-developed principles governing agroecological systems that, if violated, make systems unsustainable; that those principles can be elaborated; that once elaborated they can be converted into hypotheses appropriate to a particular agroclimatic region; that research can be designed to validate, accept, reject, modify, or develop further those hypotheses by conducting investigations and on-farm tests in the relevant regions; and that the investigations and tests can then be evaluated and interpreted in the broader context that a systems perspective provides.

The SANREM program was created to advance this process. The specific criteria for research outlined above emphasize the needs of particular agroecosystems. Regardless of the agroecosystem under investigation, however, a successful proposal within the SANREM program will have taken into consideration the following questions:

• How does the project foster conditions and a consensus for collaboration among various constituencies?

• Are collaborative and innovative research methodologies used by the project?

• Is the project interdisciplinary?

• Does the project emphasize local and traditional expertise, knowledge, and institutional development?

• Does the project address gender issues and equity considerations?

• To what extent are intended beneficiary farmers and nongovernmental organizations integrated into the design, planning, implementation, monitoring, and evaluation of the project?

• Does the project have both applied and adaptive phases to ensure that practical results accrue for resource-poor farmers within a reasonable period of time?

• Has the project established linkages with other SANREM and non-SANREM initiatives (for example, relevant CRSPs, local or regional sustainable agriculture networks and field programs, other donor activities)?