Flooding is a major issue in the Amite River watershed (3100 Km2), which drains into parts of Baton Rouge, LA. The watershed has flooded more often than any other watershed in the region. Currently, Lower Mississippi River Forecast Center (LMRFC) uses lumped parameter hydrological model (Sacramento model) for estimation of streamflow. For boundary conditions, LMRFC uses Stage III precipitation estimates (fused product of NEXRAD radar estimates with observed data), and rainfall forecasts from NOAA Hydrometeorological Prediction Center (HPC).

This study focuses on two important aspect of flood prediction: improvements in precipitation forecasts, and application of a sophisticated spatially distributed hydrologic model. We use Kalman filter formulation to improve the HPC forecasts using the Stage III estimates. Stage III data provide an ability to examine errors in the forecasts, and apply the corrections to the future forecast fields. The filtered precipitation is fed to a spatially distributed hydrologic model (Simulator for Hydrology and Energy Exchange for the Land Surface, SHEELS). SHEELS tracks water movement in multiple soil layers as well as overland flow using topography driven network of grid cells. Soil moisture accounting is driven by the energy exchange among ground, canopy and the atmosphere.

Results showing the improvements in precipitation forecasts from the Kalman Filter will be discussed in this paper. Modeled streamflow estimates from SHEELS will be compared with observed streamflow for the Tropical Storm Allison (June 2001) will also be presented.