3.2.1S
THE MAKING OF A DECISION-SUPPORT SYSTEM

PJ VERRIER

IACR-Rothamsted, Rothamsted Experimental Station, Harpenden, Hertfordshire, AL5 2JQ, UK

Background and objective
The arrival of Decision Support Systems to handle crop pest and disease problems has in the past been heralded by research workers as the answer to all prayers. They were seen as a means of delivering the science to the user enabling him to solve all those difficult problems with ease. From the farmers point of view, the idea sounded very promising - simply inform the system of the current problem and in a few moments, the perfect solution will be provided. Unfortunately the early plans did not always work as well as one would have hoped. The average crop or disease problem was not necessarily easy to model in sufficient detail to determine a good forward forecast of outcome. If the problem could be modelled well, the amount of data required to ensure an accurate assessment of the outcome was beyond the ability of the majority of farmers to provide easily. If the problem could be modelled well with easily available data, the number of factors involved in creating any form of optimisation became too great to provide a solution on the available computers. If the problem could be modelled and suitable heuristic rules be determined, a form of decision modelling could be accomplished if there was sufficient computing power available, to enable reasonable crop protection decisions to be taken. However, these solutions tended to essentially be a repetition of standard rules devised by crop advisors and applied to the currently recorded state of the crop. The farmer could learn from the presented solution - this was a good thing - but he could also determine that once these rules were learnt, the DSS was no further use.

Results and conclusions
Times have changed. The typical farmer and his advisor have access to remarkably powerful desktop or portable computers in the form of IBM PC compatible machines running the Microsoft Windows operating system. With these machines, it is possible to design a system that is intuitive to use, has access to vast amounts of data, can perform very large numbers of calculations, and can present output in a variety of ways. Because of the availability of this computing power, many of the paradigms required to build complex model based systems with highly sophisticated integrated decision analysis which can have access to large amounts of data either locally or remotely through modem high speed networks. Such systems a very complex in nature. The user is now highly computer literate, understands the modem User Interface concepts and expects very sophisticated software features to be available with any software that he is to use.

This high level complexity of (a) problem, (b) solution and (c) user interface creates the need for very sophisticated programs which can today be called true Decision Support Systems. Just what are the problems in creating such a system, what are the components that are essential and/or desirable, how do we put them together and how can we reduce the effort of production?

These questions will be addressed and concepts developed by examining the development of a number of DSS's and in particular the DESSAC generic Framework for PC based decision support systems which is being used by a number of separate groups of developers to create specific DSS's for the control of plant pest and disease.