The scope for modelling the behaviour of pollutants in the aquatic environment is now immense. In many practical applications there are effectively no computational constraints on what is possible. There is accordingly an increasing need for a set of principles of modelling that in some respects may well be different from those applicable when conceptualisation, the accuracy of the numerical solution scheme, and the inadequacies of an overly simplified model structure, were the issues of the day. Given the availability of increasingly comprehensive software, the user of a model is increasingly likely to be accelerated into a position where the issue of model calibration (identification) is an immediate problem. From the practical point of view of needing to make a decision on the control of a pollutant, the problem of identification may, or may not, be avoided. It is argued that a consistent approach to establishing whether such identification is necessary depends on establishing the significance, or otherwise, of model uncertainty. Identifying the model against field data does not have merely the goal of yielding “best” estimates of the unknown coefficients (parameters) appearing in the given model structure. It may also serve the purpose of identifying and modifying the uncertainty attaching to the model as a description of observed behaviour, which uncertainty will then be propagated forward in any predictions made with the model.

This content is only available as a PDF.
You do not currently have access to this content.