Record Display for the EPA National Library Catalog

RECORD NUMBER: 23 OF 34

Main Title Recursive Parameter Estimation of Hydrologic Models.
Author Rajaram, H. ; Georgakakos, K. P. ;
CORP Author Iowa Univ., Iowa City.;Corvallis Environmental Research Lab., OR.
Publisher c1989
Year Published 1989
Report Number EPA/600/J-89/522;
Stock Number PB91-182089
Additional Subjects Water management(Applied) ; Hydrology ; Mathematical models ; Watersheds ; Environmental effects ; Spatial distribution ; Temporal distribution ; Water pollution ; Flood forecasting ; Data quality ; Probability theory ; Case studies ; Study estimates ; Reprints ; Enhanced Trickle Down Model
Holdings
Library Call Number Additional Info Location Last
Modified
Checkout
Status
NTIS  PB91-182089 Some EPA libraries have a fiche copy filed under the call number shown. 07/26/2022
Collation 16p
Abstract
Proposed is a nonlinear filtering approach to recursive parameter estimation of conceptual watershed response models in state-space form. The conceptual model state is augmented by the vector of free parameters which are to be estimated from input-output data, and the extended Kalman filter is used to recursively estimate and predict the augmented state. The augmented model noise covariance is parameterized as the sum of two components: one due to errors in the augmented model input and another due to errors in the specification of augmented model constants that were estimated from other than input-output data. These components depend on the sensitivity of the augmented model to input and uncertain constants. Such a novel parameterization allows for non-stationary model noise statistics that are consistent with the dynamics of watershed response as they are described by the conceptual watershed response model. Prior information regarding uncertainty in input and uncertain constants in the form of degree-of-belief estimates of hydrologists can be used directly within the proposed formulation. Even though model structure errors are not explicitly parameterized in the present formulation, such errors can be identified through the examination of the one-step ahead predicted normalized residuals and the parameter traces during convergence. The formulation is exemplified by the estimation of the parameters of a conceptual hydrologic model with data from the 2.1-sq km watershed of Woods Lake located in the Adirondack Mountains of New York. (Copyright (c) 1989 by the American Geophysical Union.)