The use of Real-Time GFS and CFS Data to Improve Operational Water Resources Management in Northern California

Konstantine P. Georgakakos, Nicholas Graham, Aris P. Georgakakos*
Hydrologic Research Center, *Georgia Water Resources Institute


INFORM (Integrated Forecast and Reservoir Management) has been implemented to provide forecasts and decision support for the Northern California river and reservoir system, encompassing the Trinity River, the Sacramento River, the Feather River, the American River, the San Joaquin River, and the Sacramento-San Joaquin Delta. The implementation was undertaken in order to assist water managers with operational water management decisions under climatic and weather variability and potential change. The INFORM system integrates climate-weather-hydrology forecasting and adaptive reservoir management methods, explicitly accounting for system input and model parameter uncertainties. The system is designed to support the planning and management processes by deriving real time trade-offs among all relevant water management objectives (i.e., water supply and conservation, hydroelectric power production, flood control, and fisheries and environmental management) at user specified risk levels. System implementation is done in a distributed fashion, with links to operational forecast and management agencies in Northern California. Operational tests over an initial three-year demonstration phase evaluated by the California Energy Commission showed good operational performance both for wet and dry years. Information on INFORM is available on line at: CEC-500-2006-109.html and HRC INFORM.

The forecast component of INFORM uses three-dimensional ensemble forecasts from the NCEP Global Forecast System (GFS) out to 16 days with a dynamic downscaling method to produce ensemble forecasts of surface precipitation and temperature and of reservoir inflows. Up to the present and for longer lead times, INFORM uses ensemble forecasts of surface precipitation and temperature from the CFS and a statistical downscaling methodology to produce the aforementioned hydrologic fields for these lead times (from 1 month to 9 months). Because of the large size of the reservoirs considered in Northern California and the long response times associated with melting of the Sierra Nevada snow pack during warm episodes or during spring and summer warming, we are currently (Phase II of the project – transition to operations) examining the utility of extending the dynamic downscaling methodology beyond 16 days so that physically consistent and coherent events produced by the CFS are downscaled in a way that preserves significant precipitation or warming episodes in the watersheds of interest. Initial results of this analysis with 3-D CFS data provided by NCEP are presented, that suggest improved forecast performance out to at least 30 days with the availability of multi-level 3-D CFS data analogous to GFS data.