Agriculture Reference
In-Depth Information
2000' and 'California Pestcast'. There are also many plant and disease models for
canopy and root architecture, virtual plants, and virtual crops. Probabilistic systems
can mix elements from different systems, and there are many models built to aid
understanding of epidemics. Older established examples include Epimul, Epipre
(temporal data) and resistan (fungicide resistance dynamics). There are also many
disease/pest risk assessment programmes and libraries of electronic images to aid
identification (CABI, 'CyberPest' etc.). There are computer-based systems for
identification of organisms that use multi-access keys, for example the Biolog
bacterial identification system. Many statistical analyses require IT visualisation
tools before an understanding of the processes can be gained. For example,
techniques such as wavelet transform, degrade or enhance images to define
significant features. There are many cladistics/phylogenetic packages such as Fillip,
Genstat routines and Bionumerics. Similarly, there are many mapping packages
which allow the use of geostatistical methods for interpolation of data from sparse
samples and compositor indexing of GIS data by, for example, soil moisture,
elevation, phosphorous or nitrogen.
Some resources are freely available as teaching aids. For example 'LateBlight'
or 'Blitecast' which enable the effects of the major epidemiological parameters to
be simply simulated and represented graphically, demonstrating the principles
of epidemiology, interactively and effectively (Fry, 1990). 'DISTRAIN' is a
programme for teaching people to assess disease accurately by asking them to
estimate the percentage cover then giving feedback (Tomerlin and Howell, 1988; see
also Chapter 2).
Decision support systems highlight a problem with many IT tools. As a knowledge
delivery mechanism they can be: inflexible, prone to go rapidly out-of-date, expensive,
and often insufficiently user-friendly, and therefore frequently fail to deliver effective
solutions or live up to users' expectations (see below). Essentially, developers of IT
tools face a choice between one of two general aims in developing a new tool: (1)
they can opt for multi-objective tools which will integrate large amounts of
information, relevant to one or more important management decisions, and act as
comprehensive sources of information (traditional DSSs would fall into this
category); while such tools provide the user with potentially a single integrated
resource for helping with decision-making, they generally have high development
and maintenance requirements, for example, to keep lists of approved pesticide
products, application rates and costs up to date, (2) alternatively, they can opt for
simple tools which focus on individual decisions and utilise only (or mostly) simple
generic information; examples of this approach are disease risk predictors based on
statistical decision rules (see below, also Yuen and Hughes, 2002). The pros and
cons of this option are, roughly speaking, the inverse of those for option (1).
Decision tools of type (2) tend to provide less in the way of background or
supporting information, but have lower development and maintenance requirements
and ought to be simpler to use.
One aspect of integrating IT into the decision-making process that is sometimes
overlooked is that decisions are often made collectively by more than one person, or
after the main decision-maker has consulted several other people. These people might
collectively be termed the decision-maker's Decision Support Network (DSN). As an
