The increasing power and availability of computers has encouraged both
the development of computer-based simulation models and their use in
making decisions. It has therefore become increasingly important to un
derstand what a model is, when it can be used validly and in what circ
umstances it can be said to predict. There are also several types of m
odels not all of which are suitable for all purposes. This paper offer
s some guidelines for soil scientists. Models must be properly validat
ed, particularly when used in decisions, and a recently published vali
dation scheme is summarized. It is suggested that models should be ana
lysed not only for sensitivity to changes in their parameters but in s
ome circumstances to changes in parameter variance, particularly where
the model may not be linear with respect to its parameters. Non-linea
rity may be a problem where parameters are subjected to spatial averag
ing or interpolation. Models are sometimes developed and validated at
one scale and then used at a larger scale. Problems may result, and so
me questions are suggested that should be asked when a change of scale
is envisaged. Some of the modelling challenges implicit in the evolut
ion of sustainable forms of agriculture are discussed.