HR: 0800h
AN: H21C-0693 [Abstracts]
TI: A Hierarchical Approach to Distributed Parameter Estimation in Rainfall-Runoff Modeling
AU: * Chu, W
EM: wchu2@uci.edu
AF: Department of Civil and Environmental Engineering, University of California, Irivne, E/4130
Engineering Gateway, Irvine, CA 92697-2175, United States
AU: Gao, X
EM: gaox@uci.edu
AF: Department of Civil and Environmental Engineering, University of California, Irivne, E/4130
Engineering Gateway, Irvine, CA 92697-2175, United States
AU: Sorooshian, S
EM: soroosh@uci.edu
AF: Department of Civil and Environmental Engineering, University of California, Irivne, E/4130
Engineering Gateway, Irvine, CA 92697-2175, United States
AB:
Distributed rainfall-runoff models intend to account for the heterogeneous characteristics of rainfall distributions
and runoff generations thereby, improve the river forecast. In this study, a distributed river forecast model is built
on the hierarchy of sub-basins connected through a river–routing system. These hydrologic units (sub-basins)
possess a no-flux boundary and traditionally can be simulated by conceptual models with a limited number of
parameters. However, calibration is needed to make such a model perform well. In the case of distributed
modeling, the lack of streamflow observations inside a river system poses a challenge to estimate the model
parameters at sub-basin scales. A hierarchical approach is proposed as follows:
First, the study basin (a parent basin) is modeled in lumped mode and calibrated to obtain the optimized
parameters. In the next step, the parent basin is divided into three sub-basins (children basins). The same model
(with tripled parameters) is applied to the sub-basins driven by the rainfalls over the sub-basins and the model
parameters for each sun-basin are calibrated using the parent parameters as their prior values. After obtaining
the optimal parameters for the sub-basins, the hydrograph at the outlet of each sub-basin can be generated.
Finally, by repeating the similar procedure, each sub-basin can be taken as a parent basin and obtaining the
parameters for its children sub-basins.
Applying this method to one of the DIMP-2 test basins: the Illinois River basin at south of Siloam Spring, the
results show that (1) the streamflow results are improved by using the distributed rainfall and distributed
parameters in comparing with the lumped simulation results, and (2) taking the parent basin's parameters as the
priors can help to determine reasonable searching ranges when optimizing the parameters of children basins
and also reduce the chance of resulting in an optimum which is not physically plausible.
Applying this method to one of the DIMP-2 test basins: the Illinois River basin at south of Siloam Spring, the
results show that (1) the streamflow results are improved by using the distributed rainfall and distributed
parameters in comparing with the lumped simulation results, and (2) taking the parent basin's parameters as the
priors can help to determine reasonable searching ranges when optimizing the parameters of children basins
and also reduce the chance of resulting in an optimum which is not physically plausible.
DE: 1805 Computational hydrology
DE: 1816 Estimation and forecasting
DE: 1839 Hydrologic scaling
DE: 1846 Model calibration (3333)
DE: 1848 Monitoring networks
SC: Hydrology [H]
MN: 2007 Fall Meeting