HR: 08:15h
AN: H11B-02 [PDF]
TI: Parameter uncertainty assessment for distributed water quality models
AU: * van Griensven, A
EM: annvg@mail.,ucr.edu
AF: Environmental Sciences Department, University of California Riverside, Riverside, CA 92521 United States
AU: Meixner, T
EM: tmeixner@mial.ucr.edu
AF: Environmental Sciences Department, University of California Riverside, Riverside, CA 92521 United States
AB:
A new method is developed by the authors that assess the parameter uncertainty in distributed water quality models for
multiple outputs in an efficient and effective way. Distributed water quality models have a high number of parameters, high
parameter correlations, several output variables and a complex structure leading to multiple minima in the objective
function. General uncertainty/optimization methods based on random sampling as GLUE or local methods such as PEST are often
not applicable for theoretical or practical reasons. Therefore, the method "ParaSol" (Parameter Solutions) is developed to
perform optimization and model parameter uncertainty analysis for complex models such as distributed (water quality) models.
Optimization is achieved by adapting the SCE-UA to enable it to account for multi-objective problems while not being trapped
in a localized minimum. The simulations performed by the SCE-UA are further used for uncertainty analysis. Two methods have
been developed that select "good" parameter solutions out of the SCE-UA simulations based on an objective threshold. The
first method is based on chi-squared statistics to delineate the confidence regions around the optimum/optima. The second
method uses Bayesian statistics to define high probability regions whereby it accounts for the fact that the high probability
regions are likely to be sampled more densely.
The application of ParaSol on the flow and sediments simulations for Honey creek (Ohio) showed that, when 2 years of daily
data are used, the confidence regions are very small when the chi -squared statistics are used and even smaller when using
the Bayesian statistics.
The results were also used to validate the suitability of SCE-UA sampling for uncertainty analysis by comparing to 500000
Monte Carlo samples. It was shown that the SCE-UA sampling was more effective and efficient as none of the Monte Carlo
samples were close to the minimum or within the confidence region defined by ParaSol.
DE: 1860 Runoff and streamflow
DE: 1871 Surface water quality
DE: 6309 Decision making under uncertainty
SC: Hydrology [H]
MN: 2003 Fall Meeting