River water quality modelling - parameter uncertainty, sensitivity and estimation

Use of water quality models plays an important role in river water quality management. Through these models, the cause/effect relationships and the river assimilative capacity can be determined so that the appropriate strategies can be implemented in managing river quality. This paper describes the major components of development of a well-calibrated river water quality model concentrating on a case study on Yarra River in Victoria, (Australia). The reaction rates responsible for TKN, TN, TP and DO were considered. In order to put more effort on sensitive parameters during model calibration, an uncertainty and sensitivity analysis using Monte Carlo simulation was conducted first to identify the sensitive parameters to output water quality. A genetic algorithm optimisation procedure was then used for model calibration to estimate the reaction parameters. These parameters were later verified using flow events, which were not used in model calibration. Finally, to further-enhance confidence in water quality predictions, the one-at-a-time sensitivity analysis method was used to study the sensitivity of model parameters in the vicinity of the calibrated values. This post-calibration sensitivity analysis provides an indication of risk associated with decision making related to model predictions.