An Expert Crop Rotation Planning System (CROPS) for Implementing and Evaluating Low-input Crop and Livestock Systems

An Expert Crop Rotation Planning System (CROPS) for Implementing and Evaluating Low-input Crop and Livestock Systems

Summary

Widespread adoption of low-input, sustainable agricultural practices may be the only practical solution to the multifaceted crisis of American agriculture. Although low-input farming systems are increasingly recognized as economically viable and environmentally preferable to conventional, petrochemically based agriculture, the practical problems involved in whole-farm planning have largely not been addressed. Implementing low-input, biologically based farming systems may involve growing new crops, growing old crops in new rotation and with different tillage practices, and learning new techniques for improving soil tilth and ecological pest management. Because of the new management skills and knowledge required, the transition from conventional to low-input farming is generally perceived as an uncertain and risky venture. Furthermore, federal farm programs, and the interdependencies of farming operation often make impractical the adoption of component practices that appear attractive in isolation.

The Research and Extension project proposed here involves the use of artificial intelligence and expert systems to create a computer-based planning tool to help farmers choose whole-farm crop rotations, tillage and pest management practices that help achieve a more sustainable agriculture. Expert systems are excellent tools to deal with complex problems which require the synthesis and application of a broad knowledge base. The proposed system is based on a prototype system called CROPS (Crop Rotation Planning System). It will develop whole-farm crop rotation plans for specific crop/livestock operations. The system will then test and compare the expected economic and environmental performance of the generated plans with alternatives presented by the farmer or alternative plans generated by the system. These evaluations will be based on simulation models of whole farm economics and soil erosion.

The system will be developed in cooperation with two farmers operating diverse crop and livestock farms in the coastal plain and the Appalachian mountain areas. The farmers will provide design advice, will test the feasibility of the system on their farms, and will cooperate with a county extension agent and the project coordinators in designing a farmer training manual and workshop.

Project Results

Progress toward the tasks required to achieve each objective is described below.

Objective 1

We are using a knowledge-based approach similar to that used in some expert systems. This knowledge-based approach relies on qualitative descriptions of goals and relationships but can also incorporate quantitative information when available. Producing a working farm planning tool from the prototype requires the following steps:

Task 1: Compiling a knowledge base to represent low-input and conventional crop production strategies and practices. Knowledge-base entries describing crops in rotation schemes and under specific tillage practices have been developed for corn (for grain and silage), alfalfa, sorghum (for grain and silage), wheat grain, and barley silage. Information included are each crop's requirements for growth, time windows for planting and harvest, expected yields by soil type, and cropping sequences in which the crop can be developed. This knowledge base will be added to throughout the project.

Task 2: Developing a computer representation scheme for individual farms. A graphical and data base representation for farms has been developed which includes (a) user-defined goals, preferences, and constraints on the farming operation; (b) financial information about the farming operation; (c) livestock information: requirements for grain, silage hay, and pasture; (d) farm machinery and labor availability; (e) field maps and data on: locations, sizes, topologies, soil types, crop and pest histories.

Task 3: Developing a planning algorithm to construct potential crop rotations and practices for each field of the farm. This algorithm requires knowledge and information described in Tasks 1 and 2 to develop an overall farm plan that satisfies the requirements of the livestock operation and any financial, production, and operational constraints. The procedure is described by computer scientists as a constraint-based planning algorithm. It puts an overall plan together from smaller pieces (macro-operators), at each step checking that no constraints of the system (e.g., estimates of soil erosion for a sloped field in a corn-small grain-winter legume rotation) have been violated.

Currently, CROPS includes a constraint-based planning algorithm that allows users to specify target acreages for specific crops and ensures that soil erosion estimates are below the maximum soil erosion limits set by the Soil Conservation Service for highly erodible land. The algorithm uses the revised universal soil loss equation model, RUSLE. Work currently near completion will incorporate nutrient management considerations into the planner, as well as economic evaluations. During the 18 months we will refine the planner, incorporating more complexity and improving the system's performance.

Objective 2

The CROPS system has been constructed to provide all the needed inputs for the RUSLE model, and we have initiated the incorporation of computer code from RUSLE to enable CROPS to calculate automatically the required soil erosion parameters to estimate sheet and rill erosion based on crop rotation, soil type, field topology, and tillage practices. The knowledge-base entries defined above (Task 1) will be augmented with all the input requirements for the RUSLE model. These modifications have been designed, but are not yet implemented.

Linkage of CROPS to the WEPP (Water Erosion Prediction Project) model has also begun. WEPP is a process model of soil erosion, able to simulate the erosion or deposition occurring at any point in a field. It is coded in about 12,500 lines of Fortran code, provided to us by cooperators at the USDA National Soil Erosion Research Laboratory. By comparison, the RUSLE equation could be coded in one line. As a result, the WEPP model will be run only once, after a whole-farm plan has been generated. Execution time will therefore be increased by only a minute or so. Most of the data needed to run WEPP are contained in standard SCS soil data bases. However, additional data will have to be obtained from on-site soil samples on our cooperating farms. This will be accomplished later this year. Incorporation of the WEPP model will continue into the second year of the project, and perhaps into the third if substantial changes are made in the code.

Objective 3

Work on this objective will begin in the second year of the project. However work to develop the economic data bases has been initiated recently.

The FLIPSIM-V model developed by J. Richardson & C. Nixon (Texas A & M) will be linked to the CROPS system in the same way it was linked to the COTFLEX system developed by the PC (Stone), Richardson, and colleagues at Texas A & M University. Data describing the user's farm finances will be entered into a financial data base. Inputs required for the FLIPSIM modes include machinery complement, debt structure, labor costs, farm program information, and crop production budgets. FLIPSIM also requires information that will be contained already in the map-oriented field data base or that will be generated as part of the planning process: e.g., number and size of fields and expected yields for specific crops in specific fields.

FLIPSIM will be used to compare alternative farm-level, multi-year plans and will produce a probability estimate of farm survival, and gross and net cash farm income. FLIPSIM includes a function that ranks alternative farm plans based on the probability of receiving different levels of net cash farm income. This ranking and the program's financial comparisons of the alternative farm-level plans will be used by the program and by the user to evaluate the plans generated by the CROPS system.

Incorporation of FLIPSIM will involve initially a series of interviews with our cooperating farmers to develop datasets needed to run the model (Tasks 4 and 5). As described below, these interviews will be completed in the first year of the project.

Objective 4

Tasks 4 and 5: Two farm data bases will be developed in a series of interviews during the first year of the project to describe the two farm operations owned by the cooperating farmers. The farm data bases will include financial data as described above, as well as map-based field information, buildings and facilities, and a description of the livestock operation. An aerial photograph or similar farm map will be digitized into computer format. In some cases, soil types and topology will be determined from soil samples and surveying conducted on site.

Interviews have been conducted with one cooperator, Floyd Childress, III. His farm fields have been digitized and data on soil types have been entered into the computer. Interviews will continue this winter.

Tasks 7-8, 10-11: Evaluation of the CROPS system on the farm and its potential delivery to farmers will be accomplished through the participation of two farmers and one Extension County agent. All three will be integral members of a project design team that will meet regularly to supervise the development of the program, its interface, and to evaluate its progress and utility.

The three cooperators will be provided with a computer (MacIntosh IIsi) to run the software, and will meet during the winter of each year of the project for a demonstration and training session. A working prototype of the system will be demonstrated and subsequently delivered in year two, along with a user's guide and technical documentation. This alpha-test version will be used to detect both programming errors (bugs) and design errors that must be corrected in the next (beta-test) version, to be demonstrated and delivered in the middle of the third year. A final version will be released with revised documentation at the end of the project.

The holistic approach taken in CROPS is not new in concept, but never before has the need for the farm-level approach or the result of using it been so clearly demonstrated. We have identified a gap in the whole problem of implementing sustainable agriculture: the coordination problem that results on a whole farm from looking at environmental problems only on a field-by-field basis.

We have shown that implementing agricultural practices based on crop rotation is not merely a matter of making farm management a bit more complex. The idea that it is worth trading inputs for more intensive management is at the heart of the sustainable agriculture movement. In this case, however, we are dealing with an enormous explosion of complexity. We reduce inputs and environmental risks but are left with a truly mind-boggling problem.

The fundamental contribution of CROPS is that it, in itself, is a paradigm for the implementation of environmentally and economically sound agriculture.

This project is being continued as AS92-4.

Objectives

(1) Develop a computer-based expert system to devise whole-farm crop rotation plans and integrate low-input farming practices.

(2) Incorporate the WEPP soil erosion prediction model to analyze the effects of crop rotation plans developed in Objective 1 on soil erosion.

(3) Incorporate an economic model of a farming operation (FLIPSIM-V3) to evaluate the economic effects of potential farm plans developed in Objective 1.

(4) Evaluate the feasibility of whole-farm plans developed in Objective 1 on two Virginia crop/livestock operations.