HR: 0800h
AN: H21A-0184    [Abstracts]
TI: Diagnostic evaluation of distributed physically based model at the REW scale (THREW) using rainfall-runoff event analysis
AU: * Tian, F
EM: tianfq@uiuc.edu
AF: Tsinghua University, Department of Hydraulic Engineering, Tsinghua University, Beijing, 100084, China
AU: * Tian, F
EM: tianfq@uiuc.edu
AF: University of Illinois at Urbana-Champaign, 220 Davenport Hall, 607 S. Mathews Av., Urbana, IL 61801, United States
AU: Sivapalan, M
EM: sivapala@uiuc.edu
AF: University of Illinois at Urbana-Champaign, 220 Davenport Hall, 607 S. Mathews Av., Urbana, IL 61801, United States
AU: Li, H
EM: hli23@uiuc.edu
AF: University of Illinois at Urbana-Champaign, 220 Davenport Hall, 607 S. Mathews Av., Urbana, IL 61801, United States
AU: Hu, H
AF: Tsinghua University, Department of Hydraulic Engineering, Tsinghua University, Beijing, 100084, China
AB: The importance of diagnostic analysis of hydrological models is increasingly recognized by the scientific community (M. Sivapalan, et al., 2003; H. V. Gupta, et al., 2007). Model diagnosis refers to model structures and parameters being identified not only by statistical comparison of system state variables and outputs but also by process understanding in a specific watershed. Process understanding can be gained by the analysis of observational data and model results at the specific watershed as well as through regionalization. Although remote sensing technology can provide valuable data about the inputs, state variables, and outputs of the hydrological system, observational rainfall-runoff data still constitute the most accurate, reliable, direct, and thus a basic component of hydrology related database. One critical question in model diagnostic analysis is, therefore, what signature characteristic can we extract from rainfall and runoff data. To this date only a few studies have focused on this question, such as Merz et al. (2006) and Lana-Renault et al. (2007), still none of these studies related event analysis with model diagnosis in an explicit, rigorous, and systematic manner. Our work focuses on the identification of the dominant runoff generation mechanisms from event analysis of rainfall-runoff data, including correlation analysis and analysis of timing pattern. The correlation analysis involves the identification of the complex relationship among rainfall depth, intensity, runoff coefficient, and antecedent conditions, and the timing pattern analysis aims to identify the clustering pattern of runoff events in relation to the patterns of rainfall events. Our diagnostic analysis illustrates the changing pattern of runoff generation mechanisms in the DMIP2 test watersheds located in Oklahoma region, which is also well recognized by numerical simulations based on TsingHua Representative Elementary Watershed (THREW) model. The result suggests the usefulness of rainfall-runoff event analysis for model development as well as model diagnostics.
DE: 1804 Catchment
DE: 1846 Model calibration (3333)
SC: Hydrology [H]
MN: 2007 Fall Meeting