Abstract

Despite its enormous productivity, American agriculture is suffering serious economic problems. Many feel that these problems require better ways to bring science information and knowledge to bear on the solution of these problems. Computer facilitated simulation models and expert systems with their supporting databases represent one approach to accomplish this goal. But carrying out this new interdisciplinary research to deal with holistic systems is easier said than done. Some of the most complex biological systems exist in agriculture. This paper examines the question of what needs to be done, and suggests possible system research approaches for reaching the goal of solving the agriculture of today and the future.

Keywords: Agriculture, systems research. computer models, interdisciplinary research, collaborative research, team research. technology transfer.

1. Introduction

This paper deals with the organizational aspects of creating the necessary environment, support and structure to restructure biological information and knowledge to solve important agricultural problems. It presents possible approaches to developing, testing, and validating models, in the context of the new initiative of USDA's Agricultural Research Service to emphasize systems research using large interdisciplinary teams.

In spite of the magnitude of new data, new results, and new theories which scientists create every day, real progress is in many ways quite slow. Many think this is a consequence of the very conservative nature of the scientific establishment. It was Max Planck [13] who once wrote that

'New scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.'

This thought is often referred to as Planck's Law (as opposed to Planck's constant, although his Law is probably more fundamental than his constant).

Scientists have traditionally confined their attention primarily to a particular discipline, be it chemistry, biology, physics, agronomy, or other areas. Research is conducted in relatively narrow fields, and papers were published in a few journals in that field. The need for a change to a more interdisciplinary approach is increasingly evident. One major reason for this needed change is the vast accumulation of data and information in the scientific literature, making it difficult' to keep abreast of all research work related to one's area. This new attitude and outlook toward broader scientific activities on the part of some researchers is due to their realizing that the solution of real problems does not lie within narrow disciplines. Decreasing research budgets also tend to place increased emphasis on research that solves relevant problems. They realize they need to broaden their viewpoint and coordinate and integrate their research with that of others in other areas so as to be able to make the critical breakthroughs in truly new and innovative research.

Communications within disciplines have tended to be of the 'old boy network' type. Breaking into this network was mostly a non-technical matter. This too is changing. More journals end ' societies are being established to accommodate the larger number of researchers in scientific fields. Disciplines are being subdivided into sub-disciplines and science specialty areas. As this is done, new groups and networks are established. Some of these groups are clearly of the narrow focus kind, while a few of the others are moving in the direction of broadening their outlook and becoming the prototypes of interdisciplinary research activities.

As new technologies have developed, specifically the computer and telecommunications, new and vastly different channels of information dissemination have evolved. No longer is the printed journal the only route of 'publication'. Computer programs, computer models, expert systems, and computer readable databases are all new products of research that can be shared. These new products and mechanisms of information transfer are causing problems for those accustomed to counting journal papers to evaluate a scientist's productivity and stature. Part of the problem is, no doubt, due to difficulty in establishing who the author of such a product is, or what percentage of the research product each contributor can rightfully claim. Clearly for legal as well as professional reasons scientists should be recognized and acknowledged for their endeavors. A study (U.S. Congress [15]) released in April 1986 by the Office of Technology Assessment (OTA), the research arm of the US Congress, recognizes this problem. The report very clearly explains the problem of identifying authorship in the statement:

'A group of authors using personal computers connected by a telephone network can collaborate in writing an article, a piece of software, or a database. This work can exist in various forms, in different places, and can be modified by anyone having access to the network. This networking technology provides new opportunities to combine talents, resources, and knowledge. Also, using satellite technology, authors, scholars, or other creators from all over the world can work together simultaneously on the same project. However, this same networking capability might create problems for the intellectual property system. Copyright, for example, is granted to "original" works of "authorship". In a world where there are many authors of one work, worldwide collaboration, and ever-changing materials, a law based on the concepts of originality and authorship may become too unwieldy to administer.'

While no clear policy to deal with the problem was proposed by OTA, the fact that it is being examined at the highest levels of Government is encouraging. The first step to solving a problem is admitting or recognizing it as a problem.

The OTA study is not the only recent recognition of the different nature of scientific activities which make use of the computer as a tool. Abelson [2], writing in the American Scientist on computers and instrumentation, notes:

'The evolution of instrumentation and computers is already affecting the ways in which scientists work and the ways in which they interact. We have come a long distance from the days of sealing wax and string, when scientists constructed their own apparatus and mainly worked alone and were sole authors of papers. Industrial scientists fairly early found it desirable to work in interdiscplinary teams to reach specific goals. Such teamwork was likely to be more agreeable to gregarious people than to independent spirits, who tended to stay at universities. The large and costly facilities of big science tend to involve many people. It is not uncommon for a paper describing work in high-energy physics to have 100 or more coauthors. What has each contributed?'

This growing concern about proper recognition, both in a legal as well as a professional sense will be addressed again later in this presentation.

Systems Research

New organizational requirements

The recent rapid growth of science is slowly leading the scientific establishment and managers of science to realize that, in many ways, things are getting out of hand. All of the millions and millions of bits and pieces of scientific research that have been discovered and created have finally reached the point where many of us feel we are about to produce the straw that will break the camel's back. There is little or no organization to much of this information. More importantly, information in the form of data alone is not knowledge. Knowledge requires the existence of a theory to interrelate the interactions of separate parts of a whole organism or other system. The need to link or connect things (data and information) and make sense (knowledge) of these things is becoming clearer. The need for a matrix of such data was reported by the NIH-funded NAS study on 'Models for biomedical research - A new perspective' [12]. A similar study in the agriculture field entitled 'Enabling interdisciplinary research: Perspectives from agriculture, forestry, and home economics' [14] comes to the same conclusion. That conclusion is very simple. Components of real systems are clearly interlinked. When you act on one part of a system, you affect the other parts. The only way to manage the whole system is to understand the processes and linkages among its components.

A number of years ago a colleague who was funding a computer program to make mass spectrometry library comparisons wanted to provide additional funding to a research group to add infrared spectroscopy as a complimentary and confirming identification tool. I understand that the reason the research group gave for refusing the additional funding was that mass spectrometry was all that was needed for chemical identification. While it has taken over a decade to show that we need every clue and piece of information we can accurately obtain to identify a chemical [16,17], such narrow thinking (i.e., tunnel vision) probably still does exist in other areas. In agriculture I have found scientists who believe weeds are just another sort of plant, and don't need to be handled as an independent entity. In a purely scientific sense this is true, but as we all know, grass is preferable to dandelions in our lawns. Thus, it seems appropriate for any plant growth model to have at least two parts, one module which has information and characteristics on the plants we want to grow, and one module in which there is information and characteristics for plants (i.e., weeds) we don't want to grow. Time will tell if this is correct.

Two examples from agriculture that Illustrates the need for interdisciplinary research are conservation tillage and range management. In the case of conservation tillage management of both crop and soil systems are required to leave sufficient crop residue on the soil surface to control erosion. In the case of range management the interdisciplinary team needs to include at least the fields of plant, soil, and livestock sciences, as well as meteorology and economics to maximize output and minimize inputs.

The fact that we need interdisciplinary research does not mean that single discipline or multi-discipline research are not important parts of the problem-solving process. But these research activities need to be focused on filling specific knowledge gaps - not producing what amounts to random or non-coordinated findings. We need a new perspective that permits us to organize the vast amount of research dealing with the complex biological systems. This perspective is modeling, and it is the goal of the NIH sponsored NAS study [12] which has led to this AAAS symposium.

Implementation strategies

While it is one thing to identify the need for systems approaches, interdisciplinary research and team activities, it is another to actually do something about it. Fortunately, the Agricultural Research Service (ARS) in this need has been both identified and acted on. ARS is the in-house research arm of the US Department of Agriculture, consisting of over 8000 employees in some 140 locations, primarily in the USA. As stated [3]: 'The mission of the ARS is to plan, develop, and implement research that is designed to produce the new knowledge and technologies required to assure the continuing vitality of the Nation's food and agriculture enterprise'. ARS is clearly a problem-solving agency. The products of its research are knowledge and technology to solve these problems.

The problem of organizing research to deliver products that solve problems was explicitly addressed by the senior management of the Agricultural Research Service in the 1983 publication of the ARS Program Plan, a strategic plan for the agency for the late 1980's [4]. This plan has six objectives, including managing and conserving soil and water, maintaining and increasing crop and animal quality and productivity, maximizing domestic and export markets for agricultural products, and promoting public health and welfare through improved nutrition and family resource management. Objective 6 called for the agency to

'Develop the means for integrating scientific knowledge of agricultural production, processing, and marketing into systems that optimize resource management and facilitate transfer of technology to users.'

Central to Objective 6 is the need for modeling as stated in this ARS document:

'The modeling approach has several advantages. First, it is efficient. Time frames of many years can be simulated quickly and inexpensively for many locations and management strategies. Second, an unlimited number of management strategies can be considered, whereas in field experiments only a few can be considered. Third, modeling is a research tool that provides knowledge about agricultural processes and identifies the research that is needed to fill the gaps in knowledge. The development of models that deal with specific components of the agricultural system is described in other objectives of the ARS Program Plan. The intent of Objective 6 is to integrate those component models into models that can simulate even larger agricultural systems. Multidisciplinary teams of scientists and engineers will be needed to plan and carry out the modeling and validate the results.'

An example of the use of modeling to develop technologies that are transferrable to those who need to solve problems in agriculture is the recently established International Benchmark Sites Network for Agrotechnology (IBSNAT), funded by the US Agency for International Development(AID) [10] IBSNAT is designed to address the question of how agrotechnology can be transferred from locations where research has been performed to new locations. The desire is to integrate new locations. The desire is to integrate new crops, new products, and new agricultural practices into existing farming systems more productive without duplicating the research performed at other locations. To transfer this technology the researchers are using simulation models based on fundamental physical, chemical and biological processes. The assumption is that the processes are the same everywhere - only the environment differs. If the appropriate environmental conditions are incorporated, the model would predict equally well regardless of location.

ARS is not the only organization interested in modeling, as evidenced in a news article from Science magazine in March 1986 [9] which indicates that at NIH the National Institute for Environmental Health Sciences has been working for over five years to lay a basis for validation of in vitro tests, and the National Cancer Institute is also developing in vitro models for carcinogenesis. While some of this is in response to the Animal Rights movement, much of the motivation has come from the same needs identified by ARS.

It is interesting to note that the two major federal agencies involved in this effort of computer modeling of biological systems are ARS and NIH, both research agencies and problem-solving agencies. While the NSF is a research agency also, its goal is fundamental research and scientific knowledge for its own sake, not to feed the world, or have a healthy population. There is a need for both the market pulled research of problem-solving agencies, and the technology push of fundamental research agencies. One also should note that much of the technology push fundamental research stemming from technology push is generally oriented along disciplinary lines which follows from the fact that it is not designed to solve problems.

Following the publication of the ARS Program Plan in 1983, the initial response of researchers joining together to work towards Objective 6 was limited. Much to their credit, to give Objective 6 the required nudge, ARS management put their money where their report was and in October 1985, established the Agricultural Systems Research Institute (ASRI) and within the institute a new laboratory, the Model and Database Coordination Laboratory (MDCL).

Concurrent with the establishment of a systems research institute in ARS was the publication of a new six-year implementation plan [5] which identifies the strategies ARS intends to employ in reaching the goals stated in its Program Plan [31 The four strategies identified are shown in table 1.

All of these involve coordination and interdisciplinary research. Strategy 1, which involves technical problems critical to the agricultural sector, states that

'We have underutilized the approach of using interdisciplinary teams for problem solving. Such teams are needed to develop the agricultural management systems that can reduce production costs, maintain or improve product quality, reduce losses of products, and conserve soil and water resources.'

Strategy 2, which calls for the most effective utilization of the limited resources which are available at the Federal level, states that ARS must

'Assure coordination of similar or interrelated research projects both within ARS and between ARS and other research organizations to avoid unnecessary duplication, ensure complementarity, and speed progress.'

Further this second strategy states that ARS must

'Identify needs and allocate resources for new research approaches or new expertise (for example, computers and systems science, new techniques from biotechnology, and improved equipment and facilities) to increase the rate of progress or probability of success.'

Strategy 3 is an explicit statement of placing 'increased emphasis on using interdisciplinary teams for problem solving'. This strategy is a further realization that the problems of today and those of tomorrow are complex, and the days of uncoordinated individual research activities that do not specifically contribute to the solution of problems are, for the most part, numbered. This strategy recognizes that interdisciplinary research and coordination, while not a guarantee to success, is likely to improve the odds considerably.

Lastly, Strategy 4 is a statement of policy to

'Continue to foster the use and development of computer-based information and data-management systems and networks in ARS. Communications networks allow scientists of different disciplines and at different locations to readily share their knowledge and their research ideas and results. These same networks are used to transfer new technology to (agricultural) Extension and other user agencies and will eventually deliver information and decision support services directly to farmers and ranchers.'

Data-management systems now support the development of both modeling and expert systems in agriculture. Computer models of crop and animal production, soil erosion, and other systems are becoming useful vehicles for expressing the state of the art of many areas for scientific research. Expert systems can combine models and data from many different disciplines and then integrate the massive quantities of knowledge into manageable decision support systems for users of research. Agriculture is rapidly moving toward using both models and expert systems on the farm, in Government, and in business transactions.

The Model and Database Coordination Laboratory (MDCL) has a number of goals in the systems research area that follow directly from these four strategies. The first is to act as a national focal point in the agriculture research community to facilitate the development of interdisciplinary research teams to work on the solution of high priority problems. As part of this function, the MDCL lab will foster more and better communications within the agricultural community. To this end, the lab has become a satellite of the NIH-sponsored BIONET system 16], and has established electronic mail and bulletin board communication to improve the interaction and communication in the agriculture research community. The computer system being made available is on the Telenet, and soon will be available on the ARPAnet, and BITnet networks for ease of access.

The second goal is to evaluate and validate models, databases and expert systems in agriculture. As part of this activity, the lab is starting to develop a compute! accessible inventory database of models, databases, and expert systems to provide a clear, complete, accurate and up-to-date picture of existing, developing and planned activities in these areas. The laboratory will also work with agricultural experts to develop processes for capturing their expertise in the form of expert systems.

The third goal of the laboratory will be to include economics within the systems analyses, to provide a rational basis for deciding which new technologies the agency should pursue to have the greatest impact in solving critical national problems with limited resources.

With the laboratory only a few months old, it is too soon to tell how well these goals will be met and what adjustments will be needed to reach them. But, we have started, and hope to have as short a learning curve as possible.

To date we have initiated a number of efforts. To help MDCL serve as a national focal point for the development of interdisciplinary teams to work on high priority national needs and problems, we are focusing on two areas. One is to facilitate the efforts of existing Objective-6 teams, dealing primarily with national resources conservation and cropping systems to maximize farmer's profit while conserving resources. The other is fundamental biology using the BIONET system to link individual scientists into teams through the mechanism of shared database and software resources.

In the second area, evaluation and validation of agricultural research models and databases, the laboratory has started the laborious task of finding, organizing, and creating a computer readable inventory. The goal of this project is to collect the knowledge that already exists to help researchers, as well as administrators in determining what needs to be produced. It is hoped that as a result of this effort resources now being spent on duplicate activities can be allocated to more productive projects that fill critical knowledge gaps. Further we hope to be able to find common elements in models that can be modularized. Some modules could then be used in more than one model. For example, if an irrigation module for crop X is developed, tested and made operational, it would be very useful if it could be used for the irrigation portion of the model for crop Y. Achieving the goal of modularized models requires a common basic structure with clearly defined interface parameters. It also requires that modules be based on sound physical, chemical, or biological processes. Modularization assumes that systems are sufficiently similar and that the same processes act. Only the environment, system dimensions, and other factors which can be quantitatively measured can differ between systems having the same module. As a further party this evaluation and validation activity we hope to develop standard sensitivity analysis tools to help determine where more effort is required on the modules being developed and used. Sensitivity analysis allows one to vary each parameter in the model, one parameter at a time, and determine the relative sensitivity of model output to incremental changes in values of each parameter.

Aiding the general communication for agriculture scientists is essential to everything the lab does to facilitate the development of interdisciplinary teams. We plan to proceed first by establishing an electronic bulletin board, electronic mail, and a computer conferencing capability to improve existing communications. To this end we have initiated access to our computer systems via a number of networks, including Telenet, ARPAnet, and BITnet, working with the University of Maryland Computer Center to access the latter two networks. As we all know, communications are fundamental and critical to scientific progress. In the old days, when there was less published, less to learn and less to know, the classical means of communications were generally adequate. Those included journals, presentations and informal talks with colleagues at meetings, letter writing, and giving seminars at universities and laboratories around the country and around the world. Today the information explosion makes it impractical to cope with all of the relevant information through these means. Fortunately new technology is available to help cope with these changes. Online information in the forms of databases are readily (although perhaps not so inexpensively) available. Electronic mail, electronic bulletin boards, and computer (electronic) conferencing are being used, although not as widely as originally expected [8]. The reasons for this include the proliferation of electronic mail systems which don't talk with one another, and the lack of a critical mass of scientists in a particular area. Some of us have been using electronic mail for almost 20 years, while others have yet to use it. One clear major benefit of this new direction of organizing biological information and developing models (all of which must be computer based owing to the vast size of the project) is that those scientists who don't join the crowd are going to be left behind. Way behind! Staying at the forefront of science today requires more up-to-date information, discussions, and contacts than ever before. This need for modern communications between scientists has led the NSF to create still another network, the NSFnet [11]. The need for such a network was eloquently stated in the abstract of this multi-author (five) article:

'Scientific research has always relied on communication for gathering and providing access to data; for exchanging information; for holding discussions and meetings, and seminars; for collaborating with widely dispersed researchers; and for disseminating results. The pace and complexity of modern research, especially collaborations of researchers in different institutions, has dramatically increased scientists' communications needs. Scientists now need immediate access to data and information, to colleagues and collaborators, and to advanced computing and information services. Furthermore, to be really useful, communication facilities must be integrated with the scientists' normal day-to-day working environment. Scientists depend on computing and communication tools and are handicapped without them.'

Before leaving this topic it should be noted that one must be careful not to push too much high technology. New and strong medicine must be taken in carefully measured doses over a proper period of time. What is needed is substance, not hype. The idea that computer networks can replace existing forms of communication, collaboration, and interaction must be soundly dispelled. The personal contacts and 'office' politics that take place at meetings cannot be replaced by a computer terminal. A proper combination of computer networking and communications, along with 'classical' meetings and interactions are likely to be the best solution.

Areas of opportunities (i.e., problems)

Besides the human factors of changing attitudes and ways of conducting research, the biggest problem we see is the lack of databases and knowledge bases to develop, test, and validate expert systems Bledsoe, a pioneer in the Al field, has recently stressed this need in his presidential lecture to the AAAI community [7]. After talking about all the important Al research areas of the future, Bledsoe says 'Knowledge is still the key'.

He believes that 'a large-structured knowledge base would not only allow sharing of knowledge by numerous systems, but, if structured correctly, could provide much more robustness and functionality than is possible from number of distinct smaller knowledge bases'. He also believes that such a knowledge base must contain common sense knowledge as well as expert and encyclopedic knowledge.

The need for a 'matrix data base' as mentioned in the NlH/NAS study will take a good deal of time and effort to develop and maintain, and hopefully would be a step in the direction needed to develop the master knowledge base which Bledsoe dreams about. But almost everyone would rather develop models and expert systems, hoping the foundation and groundwork will be done by others. To change this view will require the proper management understanding of the needs of this new area of endeavor. The development of theories for whole systems requires both interdisciplinary teams as well as interdisciplinary data and knowledge. While models can and will help organize information and lead to new developments, the lack of the information and knowledge will clearly hinder practical applications of such model development. The next big step, as NIH is planning, is to refine the thoughts in the NIH/ NAS study, through a workshop and define a solid segment that establishes what is needed and how much it will cost. Then a properly funded activity to actually initiate the research and development work necessary to carry out a solid piece of information/knowledge base building must be undertaken. From there we would hope the basic framework and list of pieces of the pie that need to be put together will evolve and be accepted for further work. However, one must be careful not to just provide facilities. In a recent guest editorial in Science [1] entitled 'Computers and interdisciplinary research', Abelson lauds the material successes of the University of Illinois at Urbana-Champaign. Our question to this editorial is why is the computer resource being stressed rather than the human resource? Computers are tools to be used, and when put first they can distort the process of modeling or whatever the scientific research endeavor is. One must be very concerned about how to bring the human resource up to the speed of the computer resource. Without proper management and direction this will not occur in the near future.

Conclusion

To summarize, it is hoped that this presentation has both explained the position of, and plans of, the Agricultural Research Service in initiating new perspectives which incorporate interdisciplinary research. What is needed is the right combination of people and the right environment! It certainly promises to be an interesting couple of years ahead as ARS works toward this new area of organizing activities and undertaking research in the area of the development of new models and systems.

References

[1] Abelson, P.H., Computers and interdisciplinary research, Science, 230, Editorial (November 22,1985).

[2] Abelson, P.H., Instrumentation and Computers, American Scientist, 74 (1986)182.

[3] The mission of the Agricultural Research Service, Agricultural Research Service, US Department of Agriculture, Issued July (1982).

[4] Agricultural Research Service Program Plan, Miscellaneous publications number 1429, Agricultural Research Service, US Department of Agriculture (1983).

[5] Agricultural Research Service Program Plan, 6-year implementation plan, 1986-1992. Agricultural Research Service, US Department of Agriculture (September 1985).

[6] BIONET - National Computer Resource for Molecular Biology, NIH cooperative agreement #1-U41 RR01685 (3/1/84-2/28-89), IntelliGencties division, Intellicorp Inc., 1975 El Camino Real, Mountain View, CA 940402216 (415-%5-5575).

[7] Bledsoe, W., I had a dream: AAAI presidential address, 19 August 1985, AI Magazine 7 (1986) 57.

[8] Edersheim, P., Electronic mail hasn't delivered, but Backers say it is on the way, Wall Street Journal, July 23 (1985) 35 (E).

[9] Holden, C., A pivotal year for lab animal welfare, Science 232 (1986) 147.

[10] International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT), Contract #DAN-4054-C-207100, University of Hawaii, College of Tropical Agriculture and Human Resources, Department of Agronomy and Soil Science, Honolulu, HA 96822 (1985).

[11] Jennings, D.M., L.H. Landweber, J. H. Fuchs, D.J. Farber and W.R. Adrion, Computer networking for scientists, Science 231 (1986) 543.

[12] NAS, Models for biomedical research - A new perspective (National Academy Press, Washington, DC, 1985).

[13] Planck, M., Scientific autobiography and other papers (Williams and Norgate, London, 1950) 33-34.

[14] Russell, M.G., R.J. Sauer and J.M. Barnes, Eds., Enabling interdisciplinary research: perspectives from agriculture, forestry, and home economies, Miscellaneous publications 19-1982, Agriculture Experiment Station, University of Minnesota (1982).

[15] U.S. Congress, Office of Technology Assessment, Intellectual property rights in an age of electronics and information OTA-CIT-302, (US Government Printing Office, Washington, DC, April 1986).

[16] Wilkins, C.L., Hyphenated techniques for analysis of complex organic mixtures, Science, 222 (1983) 291.

[17] Wilkins, C.L., G. N. Giss, R. L. White, G.M. Brissey and E. Onyirinka, Mixture analysis by gas chromatography/ Fourier transform infrared spectrometry/mass spectrometry, Anal. Chem. 54 (1982) 2260.