Role of initial conditions and climate variability on seasonal streamflow forecasting in the southeastern US

Seasonal streamflow forecasts are essential in water resources management and hydroelectric power generation, especially during the extreme events. Seasonal forecasts derive their skill from slowly evolving Sea Surface Tempertaure (SST) as well as initial land surface conditions such as soil moisture and snow cover. However, uncertainty in utility of forecasts during different months, geographical regions and forecast lead times restrict application of streamflow forecasts in real time water and energy systems operation. The objectives of this study are to: 1) Evaluate benefits of using monthly updated retrospective forecasts from ECHAM4.5 General Circulation Model (GCM) for 1 to 7 months lead time in comparison to using climatology, 2) Quantify the role of initial soil moisture for different forecast lead time, and 3) Assess the role of climate variability on streamflow forecasting in the southeastern US. In this study, we will use a land surface model, statistical model and their combination. We will implement a large scale Variable Infiltration Capacity (VIC) land surface model at 1/8 degree resolution driven by downscaled precipitation from ECHAM4.5 model for 1 to 7 months lead time. The initial soil moisture conditions prior to the forecasting period will be derived by forcing the VIC model with observed meteorological data. In addition, we will develop a statistical model using retrospective precipitation forecasts from ECHAM4.5 model with lead time up to 7 months. Finally, we will use a combined approach by including VIC model simulated soil moisture as an additional predictor in our statistical model to evaluate the role of initial soil moisture. The results from all these models will be analyzed to determine the true value of using precipitation forecasts over climatology for lead time of 1 to 7 months.