HR: 1340h
AN: H23F-1687    [Abstracts]
TI: ESP forecasts for water resources: Model Uncertainty, Climate Uncertainty or Both?
AU: Park, G
EM: gihyeonp@uci.edu
AF: Center for Hydrometeorology and Remote Sensing, University of California, Irvine, E/4130 Engineering Gateway, Irvine, CA 92697-2175, United States
AU: * Imam, B
EM: bimam@uci.edu
AF: Center for Hydrometeorology and Remote Sensing, University of California, Irvine, E/4130 Engineering Gateway, Irvine, CA 92697-2175, United States
AU: Ferrer-Capdevila, M
EM: mferrerc@uci.edu
AF: Center for Hydrometeorology and Remote Sensing, University of California, Irvine, E/4130 Engineering Gateway, Irvine, CA 92697-2175, United States
AU: Sorooshian, S
EM: soroosh@uci.edu
AF: Center for Hydrometeorology and Remote Sensing, University of California, Irvine, E/4130 Engineering Gateway, Irvine, CA 92697-2175, United States
AB: Operational water resources management relies on the streamflow forecasts of reservoir inflow. In the Western U.S, where seasonal snowmelt represents the larger portion of water supplies, these forecasts are traditionally obtained through statistical regression-based estimates of April-July and water year runoff, which are then disaggregated to monthly volumes using historical relationships and forecaster judgment. An alternative approach is to use hydrologic forecasting systems, such as the West-Wide Seasonal Hydrologic Forecast System, developed by the University of Washington, to provide probabilistic forecasts in the form of ensemble streamflow predictions (ESP). Whether or not ESP forecasts are conditioned by seasonal climate forecasts, the approach places the natural variability of hydrometeorologic forcing (e.g. precipitation, temperature, snow extent¡¦) as the primary source of forecast uncertainty. This presentation will attempt to evaluate the effect of model uncertainty on the uncertainty statements issued by probabilistic forecasts generated from the California Hydrologic Forecast System (CaliForecast). CaliForecast, which is a regional implementation by the University of California, Irvine, of the west-wide forecasting system, will be used to issue ESP forecasts that account for uncertainty in model parameters. Within the system, parameter uncertainties will be assessed using the Bayesian-based Particle Filtering technique in order to obtain posteriori distributions of key model parameters for the Variable Infiltration Capacity Model (VIC-3L). The posterior distributions of model parameters, in conjunction with traditional ESP will allow the propagation of parameter uncertainty into the probabilistic streamflow forecasts. Comparison between probabilistic forecasts issued with and without parameter uncertainties will be conducted to assess the impact of parameter uncertainty for a sub-basin within the Feather River in Northern California.
DE: 1833 Hydroclimatology
DE: 1846 Model calibration (3333)
DE: 1880 Water management (6334)
DE: 3245 Probabilistic forecasting (3238)
SC: Hydrology [H]
MN: 2007 Fall Meeting