Basil Acock, Stephen R. Heller, and Stephen L. Rawlins
USDA, ARS, Systems Research Laboratory
Beltsville, Maryland, U.S.A.


Traditionally, data on crop growth are gathered, interpreted and used locally to advise farmers how to manage their crops. This work is repeated at sites all over the world. The data are published in scientific journals, but only as treatment means or as parameters for crude statistical models, and much of the information contained in the original data is lost. Collectively the data represent much more knowledge about crop growth than is available at any one location. They can be used to build and test crop simulators (models) and expert systems. However, agricultural journals will not publish complete data sets, or the full code of crop simulators or expert systems. A peer-reviewed journal to publish these data and software is needed.

Data and code are commonly stored and manipulated in electronic form, and it makes sense to communicate them in that medium. An electronic journal would be cheap, eminently searchable, and immediately available on every scientist's terminal without presenting a storage problem. It would also eliminate the time articles spend tin press.. However, there is no way of mailing an electronic article, in easily readable form, to administrators and funding agencies. The publishers must provide authors with at least one laser-printer copy of their article. In addition there are two major problems to be overcome: (1) electronic journal articles require graphics, but there is no universal standard. The journal "publishers" must find or develop their own graphics software, (2) most readers of such a journal will access it over telephone lines, a slow and uncertain means of transmission. The publishers must provide other access to the data, possibly via magnetic tape or diskettes for current articles and CD-ROMs for back issues.


Since J.B. Boussingault started field plot experimentation at Bechelbron, Alsace in about 1834 (Russell, 1961), data on mop growth have been gathered, interpreted, and used locally to advise farmers how to manage their Hops. Wherever man cultivates the soil, there are experiment stations testing various possible management strategies. The researchers gather data on the soil type, weather, management operations, crop growth and yield. Most of these data are summarized as means or are fitted to Rude statistical models and presented as parameters. This is done for two reasons. First, scientific journals will not publish large amounts of data. Second, very little of the information contained in the data is actually used in advising the farmer. Usually the farmers are just told what treatments work best under average weather conditions on the types of soils in their region. Thus, much of the information contained in the original data is lost.

Increasingly over the last 50 years, the data gathered at experimental stations have been used by plant physiologists, soil physicists and other specialists to develop an understanding of why crop plants behave as they do. In these experiments the data are effectively summarized as hypotheses about processes in the soil/plant/atmosphere system. These hypotheses are published in scientific journals, and as they are tested and become accepted, some of them are translated into practical advice for farmers. However, many of the hypotheses just languish in libraries because no one sees a practical application for them.


Now that computers are cheap, powerful and widely available, we can begin to make better use of these experiment station data and our hypotheses derived from them Crop simulators (mechanistic models) enable us to put in mathematical form our hypotheses about the processes in the plant, soil and atmosphere that affect crop growth These simulators allow us to gather up the scattered hypotheses and relate them to each other to produce a summary of all our knowledge about plant behavior. This knowledge, in the form of computer codes, becomes a tool that can be handed to farmers and other users to enable them to improve their farm management. The farmer himself or an expert system shell can run the simulator to predict the outcome of various possible management strategies and determine which is the most beneficial. In doing this, the farmer can enter into the computer, the details of the soil in his filed, his management operations and the weather experienced by the crop to date. He can then use a number of different weather scenarios for the rest of the season, to make his predictions. Used in this way, crop simulators can give specific advice for a crop growing in a certain Reid in a particular year. This is much superior to the general advice currently given by farm advisory services but it is a potential that yet has to be fully realized.

With crop simulators we can also make much better use of the data gathered at experiment stations. These data can be used to test or validate the simulator and its various component hypotheses. The data from a single experiment station only test a given crop simulator over a small pan of its possible operating range but, collectively, the data from all experiment stations working on that crop enable us to test the simulator over its entire range of conditions. This potential for using worldwide data to test and refine our understanding of plant behavior has been recognized in the International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT) project funded by the U.S. Agency for International Development. In this project, the researchers have specified the variables and format for essential input data to the crop simulators (IBSNAT, 1986). However, the soils data specified are only suitable for running one particular soil environment model. Other, more comprehensive, soil, environment models need additional data.


One of the problems with an exercise like IBSNAT is that experimenters have no real motivation for contributing to the project They must spend time collecting the data, putting them in the correct format, and sending them to the modelers, but there is little Heard for doing so. Researchers, like most laboratory animals, do the things that they are rewarded for doing. Rewards come in the form of promotion and additional research money (grants, etc.). These are awarded mostly on the basis of single authored, peer-reviewed journal articles. Posters, oral presentations, book chapters and articles in popular journals all count but are much less important The single act which would have the greatest impact in persuading researchers to publish all the data from their experiments would be to establish an electronic, peer-reviewed journal that accepts the data.


As mentioned earlier, conventional journals will not publish large amounts of data. Because of the cost and the limited number of pages available they are also unable to publish the code and documentation of crop simulators. Thus the data and the simulators needed to make full use of them are published in experiment station bulletins and other obscure places, if, indeed, they are published at all. The authors get little reward for doing this, apart from the private satisfaction of communicating their ideas.


The new journal should be electronic because this is the obvious means of storing and manipulating data and the computer code of crop simulators. Communicating electronically means that data and code would not have to be keyed into a computer again, with the ever-present possibility of introducing errors. The journal should be peer reviewed to maintain standards and to ensure that published articles count towards the author's promotion. An electronic journal would be cheap, eminently searchable and immediately available on every scientist's terminal without presenting a storage problem in his office. Because storage is not a problem, articles could be of any length, without invoking page charges. An electronic journal would also eliminate the time articles spend "in press" because the authors would, in effect, be doing their own typesetting.

Another possibility with electronic publication, that is not available in paper publication is that of continuously adding to an article after it has first appeared. We have in mind here not amending an article but instead having a file associated with it that refers the reader forward in time to subsequent discussion of the original article.

While the scope of the journal must be defined, it does not need to be too narrow because the subject matter can be subdivided and indexed without inconveniencing the readers. For instance, a plant physiologist, finding that 90% of his favorite journal deals with soil physics will soon stop subscribing but if the articles dealing with soil physics are transparent to him and he is not paying a subscription anyway, there should be no problem.


The idea of publishing an electronic journal is not new. An electronic journal Genetic Information Retrieval System. has been under consideration for several years, but publication is being delayed because the US National Libraries of Medicine is considering starting a similar journal. Recently Pergamon Press announced a new journal "Tetrahedron Computer Methodology" to publish articles on the use of computers in chemistry. This will be available in both hardcopy and electronic forms, but the latter will include supplementary material such as source code and atomic coordinates of molecules. However, the Pergamon word processing program ChemText will be needed to create and read figures in the electronic journal. Clearly there is interest in electronic journals, but there are also significant difficulties.

One difficulty is that the contents of computers do not have high visibility. Most of us pay attention to external stimuli in roughly the order: telephone calls, visitors in the office, pieces of paper on the desk, and electronic mail. An electronic journal, like electronic mail, but unlike paper journals, does not intrude into the office environment or land with a thud on the desk. An electronic journal has to be remembered and called up. For this reason, we believe it is necessary for the electronic journal to have a paper companion in the form of a quarterly listing of newly published abstracts plus editorial comments and letters. This would remind subscribers of the existence of the journal and alert them to articles of potential interest.

Another difficultly with an electronic journal is that monitor screens are not as easy to read as a page of crisp, black type, and they are much less portable. Even those of us who routinely use word-processors to edit manuscripts, usually print out a copy of the manuscript for marking up. It is easier to scan printed pages for wanted information than to scan successive screens on the monitor. For this reason, it seems likely that readers of the journal will want to print out some articles in their own offices.

There is nothing tangible about an electronic journal that can be mailed to administrators, funding agencies, and admirers. In time, the expectation of these people may change but for the moment, we have to live with an infrastructure geared to paper copies of articles. We therefore believe that we must provide each author with at least one laser printout of his article which he can photocopy for distribution at his own expense. Of course, many authors have their own laser printers and could do this for themselves. This would save the journal some work but might lead to the appearance of bogus articles that had not been accepted or even offered for publication in the journal. Some form of authentication may be necessary but we will meet that problem when it actually arises.

This immediately brings to mind another difficulty: the authors and readers are going to have many different computers and word processing programs. Fortunately, all computers handle ASCII text and most word processors can read in or dump out ASCII. The solution then is to have the entire journal in ASCII characters.

In addition to these minor difficulties there are two major problems with starting an electronic journal: graphics and access.


One major problem is that a scientific journal must handle graphs. Some ideas can only be explained with the aid of graphs and diagrams, and all arguments are easier to follow when illustrations arc used. However, there is no universal graphics standard for personal computers and the potential readers of the Journal all have different hardware and software at their disposal. We cannot expect them to buy specific pieces of hardware and software just to read the journal. At some future date, when the journal is established, this may seem worthwhile if there is some hardware/software combination which offers spectacular advantages. However, we believe that for the moment we must cater to the hardware owned by readers and supply any software needed.

Almost everybody in the U.S.A. either has, or has access to an IBM-PC or compatible machine. This, then, is our de facto hardware standard. At first we started to develop our own graphics software for the journal but soon found that there were considerable problems in working out all of the bugs, making it sufficiently user-friendly, and documenting it. Then we realized that the program Chart (3) by Microsoft could be used to generate ASCII files (called SYLK files by Microsoft). Chart is inexpensive and the authors of journal articles would have to purchase their own copy to generate the ASCII files.

These files would be part of the article and would reside in the journal along with a reconstruct-only version of Chart. Readers would download this special version of Chart, plus the files, into their own computer, reconstruct the figures and, if desired print them out. We are currently negotiating with Microsoft to provide us with this reconstruct-only version of Chart. As a backup to this, we are considering distributing the figures for each published article in hard copy in the quarterly bulletin.


A second major problem is that most readers of the proposed journal will have to access it over telephone lines which are a slow and uncertain means of transmission. This might not be a problem for reading text but could be a problem for downloading code. Errors introduced into text by noisy transmission are easily overlooked but errors in code can be disastrous. Therefore, we believe that we must also be prepared to answer requests for making copies of individual articles on diskette with the expense being borne by the requester. In any case, we plan to make "back issues" of the journal available on CD-ROM as often as warranted by the accumulation of articles. CD-ROM, Compact Disc Read Only Memory, uses the same technology developed for audio compact discs. A single S.25 inch disc holds up to 670 megabytes of information. Costs for producing the discs, which are decreasing rapidly, are now as low as $1500 for the master disc plus 100 copies. Cost per copy beyond this quantity is less than $3 each.


The idea of developing this journal was first mentioned in a letter sent to about 400 modelers last year. From those who responded, an editorial board has been assembled, consisting mostly of U.S.A scientists but with some overseas representatives. The problems of reaching overseas readers electronically are, of course, much greater than those of reaching U.S.A. readers. However, it is conceivable that if the journal does well, copies of it could be held on computers in other countries where they would be more accessible to the readers in those countries. The editorial board has decided on the aim and scope of the journal and the criteria to be used for accepting articles for publication. As soon as the agreement with Microsoft is concluded and the reconstruct-only version of Chan is received, we will commence publication. Initially, the journal will reside on a computer in the USDA, ARS, Systems Research Laboratory at Beltsville, MD, U.S.A. as a pan of the Agricultural Systems Research Resource (an electronic conference and bulletin board). However, when the journal is functioning satisfactorily, we intend to transfer it to some existing or newly created scientific society which can provide for its long-term nurture and a democratic means of changing journal policy and cycling the editorial board. Another important consideration is that a private society can collect subscription fees and charge for services; actions which are nearly impossible for a U.S. federal government laboratory.


[1] IBSNAT, Decision Support System for Agrotechnology Transfer: Documentation for IBSNAT Crop Model Input and Output Files, version 1.Q U.S. Agency for International Development and Dr. Gogo Uehara, University of Hawaii, 1986.

[2] Russell, E.W., Soil Conditions and Plant Growth, 9th ea., pp. 1688, Longmans, Green and Co., London, 1961.

[3] Chart, version 3.0. A scientific and business graphics program by Microsoft Corp., PO Box 97017, Redmond, WA 98073-9717, USA.