HR: 1340h
AN: H23F-1680 [Abstracts]
TI: The Effect of Hydrologic Model Calibration on Seasonal Streamflow Forecasts for the Western
U.S.
AU: * Shi, X
EM: xiaogang@u.washington.edu
AF: Department of Civil and Environmental Engineering
University of Washington, Wilson Ceramic Laboratory, Seattle, WA 98195, United States
AU: Wood, A W
EM: aww@hydro.washington.edu
AF: Department of Civil and Environmental Engineering
University of Washington, Wilson Ceramic Laboratory, Seattle, WA 98195, United States
AU: Lettenmaier, D P
EM: dennisl@u.washington.edu
AF: Department of Civil and Environmental Engineering
University of Washington, Wilson Ceramic Laboratory, Seattle, WA 98195, United States
AB:
Forecasts of seasonal streamflow, particularly for the spring and summer period which are dominated by
snowmelt runoff, are central to the management of the water resources infrastructure of the western U.S.
Operational approaches to seasonal streamflow forecasting, like the Ensemble Streamflow Prediction (ESP)
method used by the U.S. National Weather Service, rely heavily on manpower and/or computationally intensive
calibration of conceptual streamflow models. We suggest an alternative approach, in which a priori (e.g., based
on regional information) model parameters are used for streamflow forecasting, and a post processing bias
correction is applied using a percentile mapping approach which utilizes the past history of model errors
associated with the uncalibrated model. We evaluate intensively the impact of calibration on ESP forecasts at
eight forecast points carefully selected to span a range of basin sizes and hydroclimatic conditions across the
western U.S. At each of these sites, we apply the ESP approach, and evaluate forecast errors for a range of
forecast dates and lead times. We use both the root mean squared error (RMSE) and coefficient of prediction Cp
(which essentially is a measure of the fraction of variance explained by the forecast) to evaluate the effects of
model calibration on seasonal streamflow forecast accuracy. We find that while the bias correction approach
captures most of the accuracy achievable by model calibration, for most forecast points, forecast dates, and lead
times there remains a modest increase in forecast accuracy that can only be captured by model calibration.
DE: 1816 Estimation and forecasting
DE: 1846 Model calibration (3333)
DE: 1860 Streamflow
SC: Hydrology [H]
MN: 2007 Fall Meeting