A comparison of techniques for the estimation of model prediction uncertainty

The basic concepts of frequentist and Bayesian techniques for the identification of model parameters and the estimation of model prediction uncertainty are briefly reviewed. A simple example with synthetically generated data sets of a model for microbial substrate conversion is used as a didactical tool for analyzing strengths and weaknesses of both techniques in the context of environmental system identification. The comparison results in the practical superiority of the frequentist technique in the case of identifiable model parameters (computational efficiency). However, in the case of poor parameter identifiablility, the conceptual advantage of the estimation of parameter distributions and the use of prior knowledge make the Bayesian approach more recommendable. Because in environmental system identification prior knowledge often makes the use of overparametrized models necessary, Bayesian techniques are very important in this field and should more often be used.