Models for water quality management: the problem of structural change

This paper challenges the common belief that a calibrated and validated model is sufficient to warrant technically sound application to guide water quality planning measures. The reason is that the planning of measures intrinsically implies that the present situation is going to be changed. Consequently, the model is going to be used outside its original domain, which from a strict systems theoretic point of view is not allowed, unless experience shows that the underlying laws are universal. Due to the complexity of water systems and their contained ecology, the structure of the developed models is almost surely not universal enough to cover changed conditions, and the prediction ability will deteriorate. Various remedies to mitigate the situation are discussed, among them over-specification combined with knowledge from similar systems, cybernetic modelling, educated speculation, and flexibility by feed-back.