Sources of Parameter Uncertainty in Predicting Treatment Performance: The Case of Preozonation in Drinking Water Engineering

Abstract

This study investigates the factors that determine parameter uncertainty when applying predefined, existing models to predict the performance of a full scale treatment system from environmental engineering. The analysis is performed for ozonation of surface water, a technology applied in drinking water treatment for disinfection and oxidation of micropollutants. The pseudo first order rate constant of ozone decay k(O3) is characterized as a time dependent parameter and estimated from data obtained from three experimental setups representing upscaling stages in engineering design. To obtain meaningful uncertainty estimates, various factors need to be acknowledged: uncertainty about the model structure, uncertainty of other model parameters, uncertainty due to non-representative sampling, and errors in chemical analysis. It is concluded that an on-site automated sequencing batch reactor is best suited for estimating kinetics during operation of the full scale system. Furthermore, the transferability of information in upscaling from laboratory experiments to the full scale system is found to be critical. Although uncertainty analysis enhances the understanding of the system, it is also shown to be a subjective process depending on the knowledge and assumptions of the modeler and the availability and quality of data.