Keywords

Parameter estimation, Sensitivity analysis, Parameter uncertainty, Total uncertainty, DREAM(ZS)

Start Date

16-9-2020 2:20 PM

End Date

16-9-2020 2:40 PM

Abstract

Parameterization of water quality (WQ) models remains a challenging issue, as these models are typically characterized by a high number of parameters and long calculation time. This paper presents a solution to this problem by the use of a fast sensitivity analysis method. An uncertainty analysis is performed to evaluate the success of parameterization and to quantify the simulation uncertainty. For this purpose, we compare the results of the simple fast screening method of LH-OAT and the advanced method of PAWN using the conceptual WQ model of the River Dender, Belgium. To perform the uncertainty analysis, DREAM(ZS) is employed. LH-OAT provides the parameter ranking very similar to those of PAWN. For most WQ variables, only a few parameters are influential. Good calibration results are obtained for 4 WQ variables with only 8 parameters. A high total uncertainty indicates that a better assessment is needed on the inputs towards the model.

Stream and Session

false

Share

COinS
 
Sep 16th, 2:20 PM Sep 16th, 2:40 PM

A fast and effective parameterization of water quality models using a screening method of global sensitivity analysis

Parameterization of water quality (WQ) models remains a challenging issue, as these models are typically characterized by a high number of parameters and long calculation time. This paper presents a solution to this problem by the use of a fast sensitivity analysis method. An uncertainty analysis is performed to evaluate the success of parameterization and to quantify the simulation uncertainty. For this purpose, we compare the results of the simple fast screening method of LH-OAT and the advanced method of PAWN using the conceptual WQ model of the River Dender, Belgium. To perform the uncertainty analysis, DREAM(ZS) is employed. LH-OAT provides the parameter ranking very similar to those of PAWN. For most WQ variables, only a few parameters are influential. Good calibration results are obtained for 4 WQ variables with only 8 parameters. A high total uncertainty indicates that a better assessment is needed on the inputs towards the model.