Modelling under uncertainty: the scientific method revisited

Simle models developed from mechanistically based hypotheses and tested by simple statistical techniques, such as regression and time- series analysis, are quite useful for understanding the lumped behaviour of a complex system and even managing it. This modeling approach has been undertaken on the premise that the behaviour of complex systems is best regarded as probabilistic in nature. Furthermore, the approach can be repeated to develop more detailed models suitable for more intricate investigation of an environmental problem, in this case the control of river salinity. Part of this iterative process is shown here which allows the refinement of models by collecting new data commensurate with the new objectives.