Biological wastewater treatment model building fits and misfits

Abstract
The modeling of biological wastewater treatment processes has received much attention over the past ten years. Efforts are underway to develop a unified model for these processes which will greatly aid in the design and operation of wastewater treatment facilities. This paper presents a philosophical discussion of model building strategies augmented by a discussion of statistical problems associated with model parameter estimation and model discrimination. This discussion further illustrates with actual data, that goodness of fit is not a sufficient condition for model acceptance. Numerous rival models are examined to illustrate this point. In order to verify a model or to discriminate between rival models, they must be “Put in jeopardy.” If this attitude is not employed in model building efforts, important discrepancies in the proposed model may go undetected.