Evaluating the Effect of Data Uncertainty on the Uncertainty of Hydrologic Models
Location
ECC 216
Event Website
https://water.usu.edu/
Start Date
3-31-2008 6:35 PM
End Date
3-31-2008 6:40 PM
Description
Hydrologic modeling has been advanced over the last two decades including dramatic growths in computational power, increasing availability of distributed hydrologic observations and improved understanding of the physics and dynamics of the hydrologic system. However, the interest in the uncertainty related to the hydrologic model has increased as well, because the values predicted by hydrologic model are limited if reasonable uncertainty method is not provided. Therefore, this paper uses the 3 different hydrologic models such as SAC-SMA, SixPar and HyMod with an ensemble approach in order to study the effect of data uncertainty on the uncertainty of hydrologic models. To achieve this, we use the Multi-objective Shuffled Complex Evolution Metropolis (MOSCEM) algorithm. The MOSCEM is linked with an ensemble of perturbed input sequences to corporate the additional uncertainty due to data error represented by an additive form of an uncorrelated heteroscedastic measurement error while the mean, variance and other statistical properties of the original observational series are maintained. The procedure was tested using 40 years of historical data from the Leaf River watershed (1950 2 km ) located north of Collins, Mississippi using 50 ensemble runs for SAC-SMA and 20 ensemble runs for SixPar and HyMod models. Therefore, we obtained the optimal parameter sets and corresponding optimal objective function points for different heteroscedastic error levels: 5%, 10%, 15% and 20%. Using the optimal parameter sets and objective function points, this paper examines the effect of data uncertainty on the uncertainty of the 3 different hydrologic models as investigating the distribution of parameters and objective function with the Kernel density method and verifying the models for each error levels.
Evaluating the Effect of Data Uncertainty on the Uncertainty of Hydrologic Models
ECC 216
Hydrologic modeling has been advanced over the last two decades including dramatic growths in computational power, increasing availability of distributed hydrologic observations and improved understanding of the physics and dynamics of the hydrologic system. However, the interest in the uncertainty related to the hydrologic model has increased as well, because the values predicted by hydrologic model are limited if reasonable uncertainty method is not provided. Therefore, this paper uses the 3 different hydrologic models such as SAC-SMA, SixPar and HyMod with an ensemble approach in order to study the effect of data uncertainty on the uncertainty of hydrologic models. To achieve this, we use the Multi-objective Shuffled Complex Evolution Metropolis (MOSCEM) algorithm. The MOSCEM is linked with an ensemble of perturbed input sequences to corporate the additional uncertainty due to data error represented by an additive form of an uncorrelated heteroscedastic measurement error while the mean, variance and other statistical properties of the original observational series are maintained. The procedure was tested using 40 years of historical data from the Leaf River watershed (1950 2 km ) located north of Collins, Mississippi using 50 ensemble runs for SAC-SMA and 20 ensemble runs for SixPar and HyMod models. Therefore, we obtained the optimal parameter sets and corresponding optimal objective function points for different heteroscedastic error levels: 5%, 10%, 15% and 20%. Using the optimal parameter sets and objective function points, this paper examines the effect of data uncertainty on the uncertainty of the 3 different hydrologic models as investigating the distribution of parameters and objective function with the Kernel density method and verifying the models for each error levels.
https://digitalcommons.usu.edu/runoff/2008/Posters/10