It is also natural to ask whether prediction systems at Level C will be reliable enough for regular use. This question was dealt with at length in
Hackett (1988a), the conclusion being that use of Liebig's Law of the
Minimum [6, 7, 31] appears to lead to satisfactory results for factors which
are in demand by plants and are not at excessive levels; but when these
factors are present in excess or toxic factors are in play, the Law appears to
overestimate performance. Hackett (1988a) showed that there is abundant
opportunity here for unconventional forms of enquiry (see [23] for an
example).
To help with the evaluation of ideas put forward in this and related
publications [e.g. 15], the writer is creating a computer package for coarsely
predicting the performance of lesser-known plants by the empirical method
(Level C). Called PLANTGRO and written in BASIC, the package provides
simple methods for creating plant, soil, and climate data-files and enables
one to make predictions for any plant-soil-climate combination for which
one has made or has acquired data-files.
The policies advocated in the preceding section for such packages are
being put into practice as fully as possible. For example, all the data-files and
programs will be open to viewing, modification, and redistribution. Over
twenty factors are dealt with, and responses to seasonal variations are
simulated. Abundant information is given with each prediction to show how
it was arrived at - thus assisting checking and learning. The handbook
is being written in a style which helps non-qualified people to use the
simplest procedures without misleading themselves or others (they are then
encouraged to develop their expertise gradually to higher levels). The
registration system will be designed to bind users to each other rather than
to the makers of the package. And to help users who cannot afford formal
analysis of their soil, there is even a routine for making a first-guess soil file
from the type of verbal description of a profile that non-experts can make.
Over 60 plant files have already been made for the system (Table 3), and
it is expected that within 4-5 years of the publication of the package, users
will have collectively produced at least 1000 plant files. These early plant files
will of course be of variable standard, but by taking advantage of the quality
rating system used when making plant files, the ease with which files can be
improved and redistributed, and the ability of declared users to have the
registration records searched for the names and addresses of fellow-users
having similar interests, the quality of new plant files is expected to improve
very rapidly - certainly far quicker than conventional research systems
could achieve.
Trial disks of PLANTGRO have been in circulation for 18 months, and
it is expected that publication of the package will occur at about the same
time as this paper appears.