HR: 0800h
AN: H21C-0698    [Abstracts]
TI: Investigating the Success of Parameter Estimation Routines in Modeling Watershed Behavior under Post-fire Conditions
AU: * Jung, H Y
EM: kongh@seas.ucla.edu
AF: University of California, Los Angeles, 5731F Boelter Hall Box 951593 Los Angeles, CA 90095-1593, Los Angeles, CA 90024,
AU: Hogue, T
EM: thogue@seas.ucla.edu
AF: University of California, Los Angeles, 5731F Boelter Hall Box 951593 Los Angeles, CA 90095-1593, Los Angeles, CA 90024,
AB: Post-fire changes in land cover affect water quality as well as alter watershed flow paths. After a drastic land cover change, the parameters of a hydrological model typically change and uncertainty increases. Current operational forecasting by the National Weather Services includes use of the Sacramento Soil Moisture Accounting (SACSMA) model to predict streamflow under post-fire conditions. The goal of our work is to compare the success of various automated optimization techniques in identifying appropriate parameter sets for prediction of both pre- and post-fire watershed behavior. The successful optimization technique will be expected to select model parameters which provide not only the accurate prediction of total discharge but also predict flow from contributing sources, i.e. overland, lateral and base flow components, derived through hydrograph separation using collected geochemical data. We compare performance of the Shuffled Complex Evolution (SCE) (single and multi-step implementation), as well as the Shuffled Complex Evolution Metropolis (SCEM) algorithm coupled to the SACSMA. The pre-fire model runs using SCE-SACSMA (multi-step) and SCEM-SACSMA, using only discharge as a criterion, show reasonable simulation of total runoff and various flow components of the basin. Post-fire model simulations using the SCEM-SACSMA show less accurate simulation of total discharge and unrealistic representation of watershed behavior. Additionally, results from the SCEM parameter probability distributions reveal changes in highest probability parameter sets from pre- to post-fire periods (i.e. optimal values and related distributions are shifted). A multi-step implementation of the SCEM-SACSMA is being tested and will also be presented.
DE: 1846 Model calibration (3333)
DE: 1847 Modeling
DE: 1860 Streamflow
DE: 1873 Uncertainty assessment (3275)
DE: 1879 Watershed
SC: Hydrology [H]
MN: 2007 Fall Meeting