DECISION SUPPORT SYSTEMS FOR SUSTAINABLE AGRICULTURE 1
J. W. Jones, W. T. Bowen, W. G. Boggess, J. T. Ritchie
Univ. of Florida Univ. of Florida Univ. of Florida Michigan State Univ.
Gainesville, FL Gainesville, FL Gainesville, FL East Lansing, MI
1 Recent literature has focused attention on many of the issues that
2 jeopardize our ability to meet man's food, fuel and fiber needs for the
3 future. World population is continuing to increase, thereby creating
4 increasing demands on limited resources. Many agricultural practices threaten
5 the sustainability of our resource base and sometimes cause unacceptable
6 degradation of the environment. Clearly, agriculture is both affected by, and
7 affects, other components of our planet, hence a framework for developing or
8 improving agriculture must be based on long-term and broad-based perspectives.
9 A concept of "sustainable agriculture" is emerging as a philosophy to
10 guide agricultural research, development, and technology transfer. Embodied
11 in this concept is a recognition that agriculture is a hierarchy of systems
12 operating in and interacting with economic, ecological, social, and political
13 components of the earth. This hierarchy ranges from a field managed by a
14 single farmer to regional, national, and global scales where policies and
15 decisions are made that can influence production, resource use, economics, and
16 ecology at each of these levels. Because it is such a holistic philosophy,
17 1 Approved for publication as Journal Series No. by the Florida
18 Agricultural Experiment Station. This research was supported by the Institute
19 of Food and Agricultural Sciences, University of Florida, Gainesville, Florida,
20 and by the IBSNAT program of the US-AID, implemented by the University of Hawaii
21 under contract No. AID/DAN-4054-C-00-2071-00.
1 agricultural researchers who wish to develop sustainable agricultural systems
2 and policy makers who attempt to influence agriculture are confronted with
3 many difficulties. There are various goals embodied by the philosophy of
4 sustainable agriculture, some of which may be in conflict. Solutions that
5 work in one soil, climate, and socio-economic setting may not in others. The
6 multiplicity of problems that can prevent an agricultural system from being
7 sustainable and the many variations of practices and policies that could be
8 implemented create a situation in which rapid progress is difficult.
9 Indicators are needed to quantify the changes that occur in agricultural
10 systems over time at different hierarchical levels for studying agricultural
11 sustainability. These indicators can be grouped into four major categories:
12 productivity, socio-economic, resource, and ecology. More than one indicator
13 will be necessary to characterize the sustainability of a particular system.
14 Analysis of sustainability will then be based on measurement or prediction of
15 the long-term changes in the indicators appropriately defined for the level of
16 the system under study.
17 Suggestions on how to improve the sustainability of agricultural systems
18 include 1) reduce input use, 2) improve efficiency of resource use, and 3)
19 increase use of natural processes such as biological nitrogen fixation,
20 nutrient cycling, and integrated pest management. One approach to research is
21 to implement long-term crop management experiments with crop yields, soil
22 characteristics, environmental quality and other indicators periodically
23 measured. Although long-term studies are needed, the impracticality of
24 establishing enough field experiments to provide answers to all questions in
25 all locations should be apparent. Methods are needed, in addition to long-
26 term field experiments, to provide planners and policy makers with the
1 capability to analyze various practices and policies for improving
2 agricultural sustainability.
3 Modern computer technology can play a major role in the future
4 development of sustainable agricultural systems and in deciding on policies
5 and plans that help ensure their adoption. Decision support systems (DSS)
6 integrate data bases, models, expert systems, analysis and graphics programs,
7 and other information to assist users in planning and decision making. The
8 purposes of this paper are to present design concepts for a decision support
9 system for sustainable agriculture, to present an example of the use of an
10 existing DSS for studying sustainability issues, and to discuss enhancements
11 needed for existing models and data bases.
12 DESIGN OF A DECISION SUPPORT SYSTEM FOR SUSTAINABLE AGRICULTURE
13 In agriculture, the terminology "Decision Support System" has been used
14 to describe a wide range of computer software designed to aid various types of
15 decision makers. Such systems have been developed for making site-specific
16 recommendations for pest management (Michalski et al., 1983; Beck et al.,
17 1989), fertilizer (Yost et al., 1988), farm financial planning (Boggess et
18 al., 1989), and general crop management (Plant, 1989). These decision aids
19 have been developed for use by farmers or by extension agents who are advising
20 farmers. Many have been based on expert system concepts and the programming
21 of rules and logic used by experts in making these decisions. Users provide
22 site-specific information on crop and field conditions, then the decision
23 support system provides one or more recommendations. As such, these decision
24 support systems are usually narrow in scope and may be applicable only within
25 the region where they were developed.
1 Other decision support systems provide information and analysis
2 capabilities to assist researchers and planners in determining the best
3 management practices for various crops, soils, and climate conditions. For
4 example, the Decision Support System for Agrotechnology Transfer (DSSAT)
5 provides data bases, crop and soil models, and analysis programs integrated
6 together to evaluate weather-related risks associated with user-specified
7 combinations of crops, locations, and management practices (IBSNAT, 1986;
8 Jones, 1989). This type of decision support system has some capabilities for
9 evaluating the sustainability of certain crop systems as will be demonstrated
10 below. Existing decision support systems, however, lack the scope and
11 functionality to address the range of problems that have to be considered in
12 assessing the sustainability of agricultural systems. They have usually been
13 site oriented, focused on either production or environmental quality, but not
14 on both, and have not dealt with long term changes in soils, weather,
15 resources and economics.
16 It is impractical to attempt to design a single decision support system
17 that will address all of the issues being discussed in the sustainable
18 agriculture literature or one that will meet the goals of every researcher,
19 planner, or policy maker. However, based on the successes of recent systems,
20 such as the DSSAT, and continuing improvements in technology, it is reasonable
21 to expect that within the next few years considerable progress will be made in
22 developing relevant and useful systems to assist researchers, planners, and
23 policy makers in developing more sustainable production systems. In this
24 section, the functional requirements, components, and preliminary design of
25 decision support systems for sustainable agriculture (DSSSA) are presented.
1 Functional Requirements
2 The proposed DSSSA will have the capabilities needed by researchers,
3 planners, and policy makers to identify agricultural systems that are not
4 likely to be sustainable in a region and to analyze the potential benefits of
5 alternative practices over long periods of time. Table 1 summarizes the
6 general functional requirements of the system. It will provide access to
7 information on soils, weather, ecology, and land use and provide maps of the
8 distributions of various attributes over space in the region as well as
9 statistics on their frequencies of occurrence. For example, all saline soils
10 could be located, or all areas within one km of lakes could be identified. A
11 Geographic Information System (GIS) will form the basis for storing data in
12 the system and linking them to specific areas in the region. Through an
13 interactive user-interface, DSSSA users will be able to interactively select
14 areas for analysis within a specified region as well as the type of
15 information to display.
16 The proposed system will also be able to identify potential problem areas
17 in a region, display them for users, and categorize them as to the type and
18 probable magnitude of the problem. For example, users may wish to identify
19 all non-agricultural areas in a region that would likely be highly erodible
20 under intensive crop production practices or areas in which highly sensitive
21 or endangered species of plants and animals exist. This part of the DSSSA
22 will be based on expert system types of programs using knowledge bases
23 collected from scientists in several fields. It will compare attributes of
24 the landscape with rules and logic in the knowledge base using diagnostic
25 operations and present interpretations to users using maps, graphics, and
1 The proposed DSSSA will also contain simulation models for evaluating the
2 effects of alternative practices on various indicators of sustainability over
3 time and space. For example, it will contain crop/soil models to simulate the
4 changes in crop yield, soil organic matter, and nutrients over long periods of
5 time under specified practices. It will also contain models to simulate soil
6 erosion and nutrient and pesticide contamination of ground and surface water
7 in the region. These models will need to simulate the desired set of
8 sustainability indicators for each combination of crop, management, soil, and
9 weather condition in the region.
10 Finally, the DSSSA will provide an interactive tool for use in developing
11 plans and policies that influence the mosaic of crops and practices over a
12 region. For example, a goal may bp set by planners to produce a certain crop
13 yield in a region with a high probability level, but at the same time
14 minimizing water pollution, soil loss, and destruction of natural forests.
15 To provide these functional requirements, the DSSSA will include three
16 major classes of components: data bases, analysis tools, and planning or
17 decision aids. The data bases will store characteristics of the region under
18 study in the computer, and the analysis tools will use the data to predict
19 various sustainability indicators. The planning or decision aids will provide
20 interactive dialogue for users to access any of the data bases and analysis
21 tools and perform any of the functions in the system. Figure 1 shows a
22 schematic of the organization of the proposed DSSA with its spatial and
23 attribute data bases, analysis tools, and decision aids.
24 Data Bases
25 The DSSSA will require both spatial and attribute data bases. The
26 spatial data bases will divide the region into areas that delineate soils,
1 weather regimes, land use, socio-economic conditions, ecology, hydrology, and
2 transportation. Each of these data types are referred to as "layers" in a
3 geographic information system. By overlaying the layers, every point in the
4 region can be identified by its soil, land use, and other layer
5 characteristics. These spatial data bases provide the information for
6 creating maps of the region in which each area on the map can be colored based
7 on its characteristics to provide users with a picture of their distributions
8 over space.
9 The attribute data bases will store specific information about each type
10 of soil, land use, farm type, and management practice in the region. For
11 example, soil characteristics such as slope, color, depth, and percent sand,
12 silt, and clay by horizon would be.stored in the soil attribute file for every
13 soil in the region. The attribute files will be related to the spatial files
14 so that the spatial distribution of any attribute, such as soil slope, can be
15 plotted on maps. A GIS will provide the capabilities for storing both spatial
16 and attribute data, creating maps, and performing many calculations that will
17 be required to aggregate and summarize regional or sub-regional information.
18 Finally, a second type of attribute data base will need to store, on a
19 temporary basis, results from various models as well as plans being considered
20 for changes in land use or practices. The permanent attribute files will
21 provide inputs to run various models, and the temporary attribute files will
22 store simulated yields, soil loss, organic matter, economics, and other
23 variables which can then be displayed on maps or printed in reports. The
24 temporary attribute file will also store possible changes in land use and
25 practices being analyzed during a session.
1 Analysis Tools
2 The main analysis tools of the proposed DSSSA will be simulation models
3 and expert systems. These tools will make use of the data bases and perform
4 the various diagnostic and analysis functions listed in Table 1. The major
5 role of the expert system component will be to identify areas in a region
6 where problems are likely to exist under current or proposed practices, and to
7 suggest alternatives for additional analyses. There will be a need for
8 several simulation models in the DSSSA, particularly site specific crop-soil
9 simulation models and hydrology and water quality models. The crop-soil
10 models must include the capabilities to simulate changes in soil properties
11 for given cropping systems and the associated changes in crop production with
12 time. There are many existing crop and soil models (Jones and Ritchie, 1991),
13 but most of these currently lack the capabilities to simulate long-term
14 changes in soil properties or to simulate the effects of various soil factors
15 on growth and yield.
16 Table 2 provides a list of some of the major indicators of change for
17 studying the sustainability of cropping systems at a field scale. The crop-
18 soil models must be able to simulate these indicators in response to crop
19 rotations and fallow periods, tillage, irrigation, residue management,
20 fertilizer management, and pest management. The crop-soil models must be
21 sensitive to weather variables, simulate the uptake of nutrients and water,
22 and be sensitive to fertility and pests. Existing crop-soil models have the
23 capabilities to perform some of the analyses required for studying sustainable
24 cropping systems. An example will be presented later in which maize and
25 soybean rotations, with fallow periods between the crops, will be simulated
26 taking into account irrigation, residue management, and nitrogen fertilizer
1 management. However, existing models do not have the capability to simulate
2 all of the indicators of change at the field scale nor the crop responses to
4 Sustainability indicators at watershed or regional scales would have
5 broader scope than those in Table 2 for a field scale. For example, the
6 regional indicators would include characteristics of lakes, wetlands, and
7 natural forests. Various hydrology and water quality models exist (USDA,
8 1980; Beasley, 1977; Hornsby et al., 1990; Hahn et al., 1982). These models
9 simulate nutrient or pesticide losses from fields or watersheds to predict
10 changes in water quality and environmental contamination. By combining these
11 types of models with the cropping system models in the DSSSA, users will have
12 the capability to investigate the trade-offs between production and
13 environmental quality under alternative practices or plans.
14 Some work has been initiated in which different types of simulation
15 models are included in a regional decision support system (Lal et al., 1990).
16 This system, called the Agricultural and Environmental Geographic Information
17 System (AEGIS), uses a GIS, crop-soil models, and an erosion model to assist
18 researchers and regional agricultural planners in Puerto Rico and the
20 Economic models will also be needed to simulate the changes in farm and
21 regional economies under various scenarios. Agricultural cropping systems
22 must be profitable over time. Changes in supply and demand of inputs and
23 outputs will affect prices and thus profits. Also, changes in agricultural or
24 environmental policies can have dramatic effects on the profitability of
25 cropping systems. Thus, the DSSSA will show profits varying in response to
26 changes in prices and management practices. Some work has been initiated to
1 link farm scale economic models with crop-soil models in a GIS framework for a
2 Guatemala study of bean and maize producing areas (Thornton et al., 1990). In
3 a related study, the results of general circulation and crop-soil models have
4 been used in a spatial equilibrium model to evaluate the economic effects of
5 climate change on U.S. agriculture (Adams, et al., 1990).
6 Planning/Decision Aids
7 A collection of programs in the DSSSA will be structured in a
8 hierarchical manner, providing analysis and diagnostic capabilities at field,
9 farm, and regional scales. They will provide information to the user, allow
10 options to be selected from pop-up menus on the screen, and provide for
11 flexible map, report, and graphic presentation of results. The site or field
12 scale user interface will allow users to select a site (soil and weather),
13 specify initial conditions and management inputs, simulate the cropping system
14 over a number of years, and provide graphical and numerical outputs of
15 sustainable indicators and their variabilities. The existing DSSAT (IBSNAT,
16 1989) has many of the features needed for this field scale analysis. An
17 expansion of the capabilities of this system to analyze long-term trends can
18 form the basis for this decision aid at the field scale.
19 The regional scale planning decision aid will also provide information to
20 users and allow for options to be selected from on-screen menus, but the
21 options will be different from the field scale. Users will be able to select
22 all or part of the region based on a wide range of options, select regional
23 socio-economic characteristics, select farm and cropping system
24 characteristics and simulate over time and space the different sustainability
25 indicators. It will provide a capability for "remembering" good combinations
26 of land use practices so that users can incrementally build a plan for the
1 region with acceptable characteristics of any of the predicted attributes.
2 This framework for spatial analysis and plan building is an extension of the
3 one described by Lal et al. (1990) for a prototype regional agricultural
4 planning system (AEGIS).
5 The development of the DSSSA will require a well coordinated,
6 multidisciplinary effort. It will require definitions for all of the data
7 bases, their data structures, and methods for data collection. It will
8 require considerably more detail in its design than is presented here before
9 it can be further developed. In addition, it will require input from a wider
10 range of scientific expertise (e.g., anthropologists, wildlife scientists,
11 etc.) to design the system to ensure that appropriate knowledge, data, and
12 analysis tools are included to proyide users with credible planning tools.
13 The structure of the proposed DSSSA would lend itself for use across
14 different regions by putting in data specific to a given region. The models
15 and expert systems must then also be applicable to the region under study.
16 The models and expert systems should be tested, and the expert systems may
17 need to be revised to include regional expertise.
18 The following section describes examples of the types of analyses that
19 the crop/soil models will need to perform as a part of the overall system.
20 EXAMPLE USE OF CROP AND SOIL MODELS
21 The development of sustainable agricultural systems requires an
22 understanding of the physical, chemical, and biological responses to
23 management practices at a field production unit scale. The responses to
24 existing and proposed practices form the basis for quantitative statements
25 about the sustainability of production in the field itself and, by aggregation
1 over space, some of the sustainability indicators in a region. These
2 responses could be based on long-term experiments or on results from crop and
3 soil simulation models. Resource requirements, production, and chemical
4 losses from the fields in a region will be estimated by models and combined
5 with spatial information on socio-economics, ecology, and natural resources to
6 estimate regional sustainability indicators. Several existing models simulate
7 soil erosion and its effect on crop growth and yield. These two models
8 (Williams et al., 1984; Shaffer and Larson, 1982) also simulate the dynamics
9 of soil nitrogen and carbon under different tillage, fertilizer, and residue
10 management. The CENTURY model (Parton et al., 1987) simulates long-term
11 changes in soil organic carbon to study steady-state organic matter levels for
12 grasslands under different climatic zones. Wolf et al. (1989) developed a
13 model to study long-term crop response to fertilizer and soil nitrogen, taking
14 into account various soil, crop, weather, and management conditions. The
15 CERES-Maize model (Ritchie et al., 1989) contains a soil nitrogen and carbon
16 dynamics component described by Godwin and Vlek (1985). Therefore,
17 considerable experience has already been gained in modeling several of the
18 important changes that may occur to soils and affect their long-term
20 Because of the critical role of crop-soil models for analyses at all
21 scales in the DSSSA, two brief examples are presented here to demonstrate some
22 of the features needed by these models. In the first example, models in an
23 existing decision support system are used to estimate changes in maize yield
24 that may occur if practices result in significant soil erosion over some
25 period of time. In a second example, existing maize and soybean models are
26 combined to simulate the effects of long term crop rotations on soil organic
1 carbon and nitrogen and on crop yield under different nitrogen fertilizer
2 levels. The first example is analogous to performing 10 years of experiments
3 at a site to measure maize response to N and then repeating this experiment 50
4 to 100 years later after which time 15 cm of topsoil have been lost. The
5 second example is analogous to a long-term, continuous experiment (60 years)
6 with two crop rotations at each of 2 N levels. Crop yield is the primary
7 indicator of sustainability in both examples. In the second example, soil
8 organic carbon is also simulated to show how different practices may affect
9 this resource and affect the basic productivity of the site.
10 Soil and Climate Characteristics of the Analysis Site
11 Weather and soil data for these hypothetical analyses represent a site in
12 the Cerrado Region of Central Brazil. The climate in this region is
13 characterized by wet summers and dry winters. Annual precipitation at the
14 site averages about 1500 mm, 80% of which falls during the usual growing
15 season from November to April. Monthly mean daily temperatures range from 19
16 to 230 C at the site (Goedert, 1983).
17 Five years of weather data from a site in Cerrado (160 23' S and 490 8'
18 W) were used with the WGEN weather generator (Richardson and Wright, 1984)
19 contained in the DSSAT (IBSNAT, 1989) to simulate long-term daily weather data
20 for the models.
21 The soil used in all simulation analyses was an Oxisol clayeyy, oxidic,
22 isothermic Typic Acrustoxes), having 65% clay, 20% silt, and 15% sand
23 throughout the profile. The soil profile was 1.20 m in depth, and properties
24 used as model inputs for water holding limits, organic carbon, pH, and bulk
25 density (Table 3) were obtained from field measurements following several
26 years of cultivation (Bowen et al., 1990).
1 Simulating Yield Changes for an Eroded Soil
2 The first simulation experiment was conducted using version 2.1 of CERES-
3 Maize contained in the DSSAT (Ritchie et al., 1989). In each of 10 years, the
4 maize variety Cargill 111 was planted on November 10 at a population of 6.2
5 plants/m2. For each year, seven N fertilizer rates (0, 50, 100, 150, 200,
6 250, and 300 kg/ha) were simulated to obtain a yield response to applied N.
7 The experiment was then repeated to estimate the N response if the soil was
8 eroded. The eroded soil profile was created by assuming that 15 cm of the
9 original topsoil were removed, organic carbon was reduced in the remaining
10 soil by 11%, and soil water holding capacities of the top two layers were
11 decreased by 0.03 cm3/cm3 each, representing a decrease in available water of
12 0.9 cm.
13 Results of the ten years of simulated experiments were averaged and
14 plotted in Figure 2. When no N fertilizer was added, the yield loss was about
15 50% for the eroded soil, or about 0.65 t/ha. The response curves indicated 15
16 to 20 kg/ha of N would be needed on the eroded soil to produce the same yield
17 as no N on the non-eroded soil. Under high levels of N, the yield decrease
18 expected due to changes in soil properties by erosion remained about 0.65
19 t/ha. However, due to the asymptotic yield responses at these high N levels,
20 the amount of extra N needed by the eroded soil to offset erosion increased to
21 40 to 60 kg/ha. The lower yields of the eroded soil in this example were due
22 to decreased N supply from organic matter in the soil as well as some increase
23 in water stress due to its lower water holding capacity.
24 Depending on the practices of the farmer, the magnitude of expected
25 losses in this hypothetical example may or may not be of major consequence.
26 If the normal practice is to apply 200 kg N/ha, then the yield loss would be
1 about 6.5%. By increasing N to 250 kg/ha for the eroded soil, yield could be
2 maintained. However, the resource use efficiency (yield per unit of N in this
3 case) would be lower. Furthermore, the cost of N may prevent the use of
4 additional N to maintain yield.
5 If the farmer did not use N, the 50% loss in yield could be devastating,
6 and it may not be possible to add 15 to 20 kg N to maintain the yield due to
7 lack of availability or high cost of N. Clearly, the magnitude of the impact
8 of this loss in soil productivity would depend upon the socio-economic
9 setting. Results from this field scale simulation provide information for
10 computing productivity, resource, and socio-economic indicators of
11 sustainability at the field scale.
12 Simulating Losses in Soil Organic Matter and Crop Yield
13 In contrast to the previous example, this study simulates the long-term
14 effect of various practices on soil properties fr v-.-u~a praeet-eeaz as well
15 as crop yield. In particular, losses in soil organic matter are simulated for
16 60-year sequences of cropping practices.
17 Maize (CERES-Maize V2.1) and soybean (SOYGRO V5.42) models were modified
18 to simulate long-term sequences of maize-fallow and maize-fallow-soybean-
19 fallow crop rotations under two nitrogen levels (0 and 150 kg N/ha). The
20 soybean model was modified by Hoogenboom et al. (1990) to simulate soil N
21 processes, using submodels from CERES-Maize, and symbiotic N fixation. The
22 soil N models in both CERES-Maize and SOYGRO were modified by incorporating
23 the method reported by Parton et al. (1987) to represent soil C and N pools
24 and their decomposition rates. Fresh organic matter dynamics as well as
25 inorganic N transformation, transport and uptake remained as they were in the
1 CERES-Maize model. The revised maize and soybean models were then modified to
2 run in sequence with fallow periods (bare soil) between crop seasons.
3 The treatments were crop rotation (maize each year or soybean-maize, each
4 with bare soil fallow during the dry season) and N fertilizer level for the
5 maize crop (0 or 150 kg N/ha). The maize variety, row and plant spacings, and
6 planting dates were the same as for the previous example. A tropical soybean
7 variety, Jupiter, was planted each year on November 10 in 0.9 m rows at a
8 density of 30 plants/m2. Residue from each crop was assumed to be
9 incorporated to 10 cm after harvest. The biomass residue and the C:N ratio
10 depended on growth during each year. Neither of the crops was irrigated in
11 any of the rotations.
12 Figure 3 shows the changes in~maize yields for the different rotations
13 and N treatments over the 60 years of simulated experiments. Year-to-year
14 variability in each treatment was due to differences in weather in each of the
15 years. In the fertilized treatments, maize yields ranged from about 5.5 to
16 12.2 T/ha, with only a small effect of soybean in the rotation. There was no
17 apparent trend in this sequence of maize yields under high N levels even
18 though soil organic carbon declined during this time period (Fig. 4). The
19 annual variability masked any apparent trends. However, additional analyses
20 similar to the first example may show some loss in productivity.
21 Results for both crop rotations without N fertilizer showed significant
22 yield declines during the first 10 to 15 years in addition to annual
23 variability. Without N, the yields of the maize-fallow sequence declined from
24 an average of about 1.0 T/ha over the first three years to no significant
25 yield after 15 years. This decline was attributed directly to the loss in
26 soil organic carbon (Fig. 4). Over this same period, maize yield in the
1 maize-fallow-soybean-fallow rotation treatment without N fertilizer declined,
2 but remained about 0.5 to 1.0 T/ha above the yield of the maize-fallow
4 Organic carbon dropped for all treatment combinations (Fig. 4), but more
5 so for the treatments without inorganic N fertilizer. Crop residue added back
6 to the soil was considerably higher in the fertilized treatments, resulting in
7 a less rapid decline. Such declines in soil organic carbon would not be
8 expected to be continuous and equal under all circumstances. Other
9 simulations with these and other models (Parton et al., 1987; Wolf et al.,
10 1989) show that soil organic carbon reaches steady state levels which depend
11 on management practices, soil, and climate.
12 Results from this study estimate that a decline in productivity by maize
13 will occur rather quickly without the replacement of N in this soil. Rotation
14 with soybean supplied some N to sustain productivity. Further improvements in
15 management could be investigated, such as timing of soybean residue
16 incorporation or growing green manure crops prior to each maize crop as shown
17 by Bowen et al. (1990). Other benefits of rotation may occur in cropping
18 systems but were not considered in this example.
19 Crop-Soil Modeling Needs
20 For use in studying sustainable agriculture issues, crop models must
21 respond to a wide range of soil conditions, such as water, nutrients, organic
22 matter, acidity, salinity, and soil organisms. Currently, most crop models
23 respond to weather and soil water, and some respond to soil N (Jones and
24 Ritchie, 1991). Research is needed to better understand the effects of other
25 soil conditions and how the plant responds to various combinations of stress
26 conditions. Understanding the mechanisms of genetic differences in responses
1 is also needed to simulate differences in cultivars in attempts to develop
2 sustainable crop systems. In addition, more work is needed to incorporate
3 these responses into existing crop models and to test them under a wide range
4 of conditions.
5 Within a single crop season, changes in many soil properties are small
6 enough to ignore them for predicting crop responses to soil, weather, and
7 management over a season. Analysis of sustainable cropping systems, however,
8 will require that the soil models simulate the slowly changing properties as
9 well. Changes in soil organic carbon, soil depth, salinity and other
10 variables must be included, and they must be coupled to crop models to
11 properly account for the feedback effects between these two components. These
12 models must also be tested under a,range of conditions and over long time
13 periods. It is for this reason that long-term experiments are so important.
14 Monitoring of these long-term experiments should include a minimum data set
15 required to run crop and soil models and to compare experimental and simulated
16 indicators of sustainability.
18 In this chapter, a systems approach to study sustainability is proposed.
19 This requires an explicit definition of the components to be considered, their
20 interrelationships, variables that characterize their response to practices
21 and environment, and external variables that act to influence the system under
22 study. To illustrate this approach, emphasis was placed on an agricultural
23 field, and indicators of the sustainability of a field production system were
24 proposed. Crop and soil simulation models were used to show how
25 sustainability at this scale would be affected by practices that contribute to
1 soil erosion and soil organic matter loss. A point was made that existing
2 crop/soil models need to be expanded to include other factors for broad
3 application in a DSSSA, and that these models need to be tested over time and
5 This is only a starting point. The forces that prevent agricultural
6 sustainability extend beyond the soil and climate changes that may occur at a
7 field site. We do not live in a static world. Changes in populations,
8 resource availability, economics, and policies at regional, national, and
9 international scales create forces for change in agriculture. Therefore,
10 planners and policy makers must be able to respond to these changes and make
11 decisions that facilitate the development of sustainable agricultural systems
12 that adapt to the changes. A systems approach is needed to assist decision
13 makers understand the possible changes in agriculture due to changes that are
14 occurring or are likely to occur in socio-economic, ecological, and political
15 conditions and to allow them to evaluate different policies and plans.
16 Because the forces of change are variable over time and space, analysis tools
17 in a DSSSA must allow for complete specification of these variables to create
18 the appropriate setting in which analyses are to be performed.
19 Development of a DSSSA is not a trivial task. It will require a long-
20 term commitment by interdisciplinary research teams to fully design the system
21 and its needed models, data bases, and other analysis tools and to implement
22 and test the various components. It will require an international effort in
23 which scientists from different institutions and disciplines cooperate in the
24 creation of such a system. The logistics of the communications required for
25 accomplishing this goal far exceed those of any previous agricultural research
1 The payoffs of a large scale effort to develop a DSSSA would more than
2 justify the efforts and costs involved. It would provide a scientific
3 framework for learning how to manage agricultural production in the larger
4 contexts of its interactions with other components of the earth. It would
5 provide context-sensitive analysis tools for planners and policy makers in
6 different socio-economic, political, agricultural, and ecological settings to
7 help them understand the many consequences of their decisions. It would also
8 provide for improved education across scientific disciplines, the general
9 public, farmers, and policy makers.
2 Adams, R. M., C. Rosenzweig, R. M. Peart, J. T. Ritchie, B. A. McCarl, J. D.
3 Glyer, R. B. Curry, J. W. Jones, K. J. Boote, and L. H. Allen. 1990.
4 Global climate change and U.S. agriculture. Nature 345:219-224.
5 Beasley, D. B. 1977. ANSWERS: A mathematical model for simulating the
6 effects of land use and management on water quality. Ph.D. Thesis.
7 Agricultural Engineering Department, Purdue University, West Lafayette,
9 Beck, H., Jones, P. H., and Jones, J. W. 1989. SOYBUG: An expert system for
10 soybean insect pest management. Ag. Systems 30(3):269-86.
11 Boggess, W. G., van Blokland, P. J,., and Moss, S. D. 1989. FinARS: A
12 financial analysis review expert system. Ag. Systems 31:19-34.
13 Bowen, W. T., J. W. Jones, and U. Singh. 1990. Simulation of green manure
14 nitrogen availability to subsequent maize crops. Agronomy Abstracts.
15 Am. Soc. of Agronomy, Madison, WI. p. 14.
16 Godwin, D. C. and P. L. G. Vlek. 1985. Simulation of nitrogen dynamics in
17 wheat cropping systems. In W. Day and R. K. Atkin (eds,), Wheat Growth
18 and Modeling. Plenum Press, New York.
19 Goedert, W. J. 1983. Management of the Cerrado soils of Brazil: A review.
20 J. Soil Sci. 34:405-428.
21 Hahn, C. T., H. P. Johnson, and D. L. Brakensilk (eds.). 1982. Hydrologic
22 Modeling of Small Watersheds. Amer. Society of Agr. Engineers, St.
23 Joseph, MI 49085.
:z.- .. ^ "
1 Hoogenboom, G., J. W. Jones, and K. J. Boote. 1989. Nitrogen fixation,
2 uptake and remobilization in legumes: A modeling approach. Agronomy
3 Abstracts. American Soc. of Agronomy, Madison, WI p. 16.
4 Hornsby, A. G., P. S. C. Rao, J. G. Breath, P. V. Rao, K. D. Pennell, R. E.
5 Jessup, and G. D. Means. 1990. Evaluation of models for predicting fate
6 of pesticides. Project Report. DER Contract No. WM-255, Soil Science
7 Department, University of Florida, Gainesville, FL 32611.
8 IBSNAT. 1986. Decision support system for agrotechnology transfer.
9 Agrotechnology Transfer 2:1-5. Department of Agronomy and Soils,
10 University of Hawaii, Honolulu, HI 96822.
11 IBSNAT. 1989. Decision Support System for Agrotechnology Transfer (DSSAT)
12 Version 2.1 Dept. of Agronomy and Soils, University of Hawaii, Honolulu,
13 HI 96822.
14 Jones, J. W. 1989. Integrating models with data bases and expert systems for
15 decision making. IN Climate and Agriculture: Systems Approaches to
16 Decision Making. A. Weiss (ed.). University of Nebraska, Lincoln, NE.
17 pp. 194-211.
18 Jones, J. W. and J. T. Ritchie. 1991. The use of crop models in irrigation
19 management IN Management of Farm Irrigation Systems. G. J. Hoffman,
20 T. Howell, and K. H. Solomon (eds.). ASAE, St. Joseph, MI (In Press).
21 Lal, H., J. W. Jones, F. H. Beinroth and R. M. Peart. 1990. Regional
22 agricultural planning information and decision support system. Proc. of
23 Conference on Application of GIS Simulation Models and Knowledge-Based
24 Systems for Land Use Management, VPI and State University, Blacksburg,
25 VA. Pg. 51-60.
1 Michalski, R. S., J. H. Davis, V. S. Visht and J. B. Sinclair. 1983. A
2 computer-based advisory system for diagnosing soybean diseases in
3 Illinois. Plant Disease 67:459-463.
4 Parton, W. J., D. S. Schimel, C. V. Cole, and D. S. Ojima. 1987. Analysis of
5 factors controlling soil organic matter levels in Great Plains
6 grasslands. Soil Sci. Soc. Am. J. 51:1173-1179.
7 Plant, R. E. 1989. An integrated expert decision support system for
8 agricultural management. Ag. Systems 29:49-66.
9 Ritchie, J. T., U. Singh, D. Godwin, and L. Hunt. 1989. A user's guide to
10 CERES-Maize-V2.10. International Fertilizer Development Center, Muscle
11 Shoals, Alabama 35662.
12 Shaffer, M. J. and W. E. Larson (ed.). 1982. Nitrogen-tillage-residue
13 management (NTRM) model: Technical documentation. Research Report.
14 USDA-ARS, St. Paul, MN.
15 Thornton, P. K. and G. Hoogenboom. 1990. Agrotechnology transfer using
16 biological and socio-economic modeling in Guatemala. Research Report.
17 The Edinburgh School of Agriculture, Edinburgh, UK. 40 pp.
18 USDA. 1980. CREAMS: A field-scale model for chemicals, runoff, and erosion
19 from agricultural management systems. IN W. G. Knisel (ed.),
20 Conservation Research Report 26., USDA, Washington, DC.
21 Williams, J. R., C. A. Jones, and P. T. Dyke. 1984. A modeling approach to
22 determining the relationship between erosion and soil productivity.
23 Trans. Am. Soc. Agric. Eng. 27:129-144.
1 Wolf, J., C. T. de Wit, and H. van Keulen. 1989. Modeling long-term crop
2 response to fertilizer and soil nitrogen. I. Model description and
3 application. Plant and Soil 120:11-22.
4 Yost, R. S., G. Uehara, M. Wade, M. Sudjadi, I. P. G. Widjaja-Adhi and Zhi-
5 Cheng Li. 1988. Expert systems in agriculture: Determining lime
6 recommendations for soils of the humid tropics. Research Extension
7 Series 089-03.88 College of Tropical Agriculture and Human Resources,
8 University of Hawaii, Honolulu, HI. 8 pp.
Table 1. General functional requirements of a Decision Support System for
Sustainable Agriculture for use by researchers, planners, and policy
1. Provide information to users on characteristics of a region under study,
such as land use, soil types, weather characteristics, farm types, and
2. Assist users in identifying areas in which existing production systems
are likely to be non-sustainable.
3. Suggest changes in practices that should be considered in fragile
agroecosystems to improve the sustainability of agricultural production
systems, natural resources, and the environment.
4. Analyze alternative practices to determine optimal practices to meet
short and long-term production, economic and environmental goals for a
site or farm within the region.
5. Provide policy makers with a consequence analysis tool for evaluating the
probable impacts of various policies and practices on productivity,
stability and sustainability of agriculture in a region.
6. Provide methods for developing regional agricultural plans that meet
desired economic, production, and environmental goals.
Table 2. Indicators of changes in field-level cropping systems that may be
useful in.studying agricultural sustainability.
Productivity Socio-Economic Resource Ecology
Yield Cost of Production Chemical Use Chemical Losses
Yield Quality Net Returns Water Use Water Quality
Yield Stability Soil Beneficial Organisms
Soil Properties Soil Toxins"
Table 3. Soil water holding characteristics (LL and ODUL, lower and drained
upper limits, respectively), bulk density, pH, and initial soil
organic carbon for the Oxisol from Brazil used in the simulations
(from Bowen, 1990).
Depth OLL BOL Density Carbon pH
0-15 0.170 0.250 0.93 1.70 5.6
15-30 0.237 0.337 1.03 1.35 5.4
30-45 0.250 0.330 1.01 1.10 5.3
45-60 0.256 0.329 0.95 0.93 5.2
60-75 0.259 0.325 0.93 0.81 5.1
75-90 0.250 0.319 0.91 -0.76 4.9
90-105 0.250 0.322 0.91 0.69 5.0
105-120 0.250 0.326 0.91 0.63 5.0
LIST OF FIGURES
Schematic of data bases, models, and analysis tools for the
proposed Decision Support System for Sustainable Agriculture.
Simulated maize yield response to nitrogen fertilizer application
for the Cerrado Region of Brazil for the original soil (M) and
the same soil after 15 cm of topsoil erosion (A).
Simulated maize yields for the Cerrado Region of Brazil on an
oxisol over 60 years. Treatments were maize-fallow with 150
kgN/ha, 9 maize-fallow with no N applied, A soybean-fallow-
maize-fallow with 150 kgN/ha on maize, and soybean-fallow-
maize-fallow with no N applied.
Simulated organic carbon for the oxisol soil profile for 60 years
in the Cerrado Region of Brazil. (See Figure 3 for symbols
..... .' '-
0 4 I
0 50 100 150 200 250 300
INORGANIC N APPLIED (kg/ha)
10 20 30 40 50
8 5 -- A -. ,
0 0 "
Z "-L -.
W 80 A *<
0 10 20 30 40 50 60