HR: 0800h
AN: H21C-0692 [Abstracts]
TI: Performance and Probabilistic Verification of Regional Parameter Estimates for Conceptual Rainfall-runoff Models
AU: Franz, K
EM: kfranz@iastate.edu
AF: Iowa State University, Geological and Atmospheric Sciences
3023 Agronomy Hall, Ames, IA 50011, United States
AU: * Hogue, T
EM: thogue@seas.ucla.edu
AF: UCLA, Department of Civil and Environmental Engineering
5732 Boelter Hall, Los Angeles, CA 90095-1593, United States
AU: Barco, J
EM: ojbarco@ucla.edu
AF: UCLA, Department of Civil and Environmental Engineering
5732 Boelter Hall, Los Angeles, CA 90095-1593, United States
AB:
Identification of appropriate parameter sets for simulation of streamflow in ungauged basins has become a
significant challenge for both operational and research hydrologists. This is especially difficult in the case of
conceptual models, when model parameters typically must be "calibrated" or adjusted to match streamflow
conditions in specific systems (i.e. some of the parameters are not directly observable). This paper addresses
the performance and uncertainty associated with transferring conceptual rainfall-runoff model parameters
between basins within large-scale ecoregions. We use the National Weather Service's (NWS) operational
hydrologic model, the SACramento Soil Moisture Accounting (SAC-SMA) model. A Multi-Step Automatic Calibration
Scheme (MACS), using the Shuffle Complex Evolution (SCE), is used to optimize SAC-SMA parameters for a
group of watersheds with extensive hydrologic records from the Model Parameter Estimation Experiment
(MOPEX) database. We then explore "hydroclimatic" relationships between basins to facilitate regionalization of
parameters for an established ecoregion in the southeastern United States. The impact of regionalized
parameters is evaluated via standard model performance statistics as well as through generation of hindcasts
and probabilistic verification procedures to evaluate streamflow forecast skill. Preliminary results show
climatology ("climate neighbor") to be a better indicator of transferability than physical similarities or proximity
("nearest neighbor"). The mean and median of all the parameters within the ecoregion are the poorest choice for
the ungauged basin. The choice of regionalized parameter set affected the skill of the ensemble streamflow
hindcasts, however, all parameter sets show little skill in forecasts after five weeks (i.e. climatology is as good an
indicator of future streamflows). In addition, the optimum parameter set changed seasonally, with the "nearest
neighbor" showing the highest skill in the winter months and the "climate neighbor" showing the highest skill in
the summer months.
DE: 1816 Estimation and forecasting
DE: 1846 Model calibration (3333)
DE: 1873 Uncertainty assessment (3275)
DE: 1874 Ungaged basins
SC: Hydrology [H]
MN: 2007 Fall Meeting