We will present streamflow simulations from short-term forecasts, a seasonal hindcast, and a seasonal forecast for the 1997-1998 winter using the Regional Climate System Model (RCSM). We will discuss the results, applications, improvements and future directions for basin-scale coupled streamflow predictions.
The coupled atmosphere-streamflow simulation uses a semi-distributed hydrologic model (TOPMODEL) and a lumped hydrologic model (Sacramento) nested in the Mesoscale Atmospheric Simulation (MAS) model and the Soil-Plant-Snow (SPS) model. The MAS-SPS model provides forcing to TOPMODEL and Sacramento such as precipitation, snowmelt, low-level air temperature and radiation. This coupled atmosphere-streamflow part of the RCSM has been employed in weather and streamflow predictions and seasonal hydroclimate studies for the southwestern United States during the past five years.
Automated 48-hr streamflow predictions start from the NCEP AVN forecast that provides initial and lateral boundary conditions to the atmosphere-hydrology part of the RCSM. The MAS model produces regional-scale predictions at a 20km resolution using the large-scale forecast data and provides basin-average forcing data for TOPMODEL. Streamflow predictions were produced at 6-hr time steps for two sub-basins of the Russian River at the northern California Coastal Range. Streamflow forecasts showed good skill during periods of continuous precipitation when uncertainties in the antecedent soil moisture conditions were small. Streamflow forecast skill was fair to poor during dry periods when base flow is a dominant source of streamflow. As a result, this coupled streamflow prediction technique was especially useful in predicting the timing and amount of flow during heavy precipitation events which are important for flood forecasting and reservoir operations.
A seasonal hindcast experiment utilized the initial fields of the NCEP AVN forecast to run the RCSM. The main purpose of this experiment includes an evaluation of the RCSM and a water resources assessment. We have obtained a good agreement between the simulated and observed streamflow over the two sub-basins, the Hopland and the Healdsburg basins, of the Russian River with correlation values of 0.61 and 0.81, respectively. Both TOPMODEL and the Sacramento model captured the peak streamflow well.
A seasonal hydroclimate prediction experiment took place for the period 1 November 1997 to 31 March 1998 as a collaboration between University of California and the NCEP. The large-scale forcing data was produced by the UCLA GCM driven by the SST forecast from NCEP. The main focus of this experiment is to establish a seasonal prediction system for the western United States, examine the performance of the system, and develop a method for simulating and producing information useful for water resources management and long-term planning. The strong ENSO during the 1997-1998 winter season made this study more important. We will present the predicted streamflow characteristics (e.g., frequency and duration of high flow events, monthly discharge volume) at the two basins described above. As expected, daily time series generated from the GCM-driven seasonal forecast is not useful at this early stage. These studies indicate that this approach may be useful for determining the probability of high streamflow during a winter season. Further refinement of the large-scale (UCLA) model performance, improved nesting methods, and statistical evaluation of the seasonal predictions are required before this approach is applicable