HR: 0800h
AN: H21C-0691    [Abstracts]
TI: A signature index approach to diagnostic evaluation and parameter estimation of watershed models
AU: * Yilmaz, K K
EM: koray@hwr.arizona.edu
AF: SAHRA Department of Hydrology and Water Resources, University of Arizona, 1133 E. James E Roger Way, Tucson, AZ 85721, United States
AU: Gupta, H V
EM: hoshin.gupta@hwr.arizona.edu
AF: SAHRA Department of Hydrology and Water Resources, University of Arizona, 1133 E. James E Roger Way, Tucson, AZ 85721, United States
AU: Wagener, T
EM: thorsten@engr.psu.edu
AF: Department of Civil & Environmental Engineering Pennsylvania State University, 226b Sackett Building, University Park, PA 16802, United States
AB: With ever increasing complexity of hydrologic models the use of regression based aggregate criteria in the model identification process increasingly limits our ability to identify model components that properly represent the actual hydrologic processes within the watershed and those that do not. The major limitation is that these measures do not provide a strong basis for detecting the causes and resolution of model performance inadequacies. Our presentation will demonstrate a signature index approach in an effort to bring in more hydrological understanding in model evaluation and parameter estimation problems. The approach starts with identification of primary functions of any watershed/model at a hierarchy of timescales and continues with the formulation of signature indices, derived from input-state-output behavior of the watershed, which can be used for proper representation of these functions in the watershed model. Preliminary analysis using the Sacramento Model showed that the approach satisfactorily identifies model parameter/signature index relations; however, secondary effects were unavoidable due to parameter interactions. The solution to this issue is to apply the approach to more parsimonious models first –with fewer interacting parameters- and then increase the model complexity in a stepwise fashion while formulating new signatures if necessary. A Monte Carlo constraining approach is utilized to identify model parameters resulting in improved signature index match when compared to a baseline model run. The advantage of the approach is its effectiveness and efficiency in extracting hydrologically relevant information from observations, which in turn increases its diagnostic power and makes it useful for watersheds with limited historical observations.
DE: 1816 Estimation and forecasting
DE: 1846 Model calibration (3333)
DE: 1847 Modeling
SC: Hydrology [H]
MN: 2007 Fall Meeting