Provision of the Optimum Flood Prediction Accuracy and Lead-time

The preparation of extreme precipitation and flooding event requires sufficient prediction accuracy and lead-time, because it is highly related to the human life and economical effects. However, many current studies may not satisfy both of requirements sufficiently because of time limited information such as direct measurements from precipitation gauges, and radar data. Presumably modeling forecasting of precipitation based on satellite images can be used for providing better conditions. Generally, the accuracy of flood prediction relies on the estimated precipitation and streamflow information. In this study, the former component of flood prediction is investigated. The WRF 3.0 (Weather Research Forecasting) model with hydrologic models (e.g. HEC-HMS (Hydrologic Engineering Center - Hydrologic Modeling System) was applied for historical hurricane events such as Hurricane Floyd 1999 over Tar-Pamilco watershed (Approximately 5500 square miles) in North Carolina. Global forecast system model (GFS) is used as initial meteorological data set. For accuracy evaluation, precipitation data from different sources (e.g. rain gauges, next generation radar (NEXRAD), and WRF) are adopted into hydrologic model as a meteorological component and compared with USGS rainfall discharge observations. In the WRF simulation, nested size of domain (4 km X 4 km) is used for finer resolution of rainfall data. For hydrological model application we used the controlled conditions (without parameter changes with different meteorological information). Rain gauge data and NEXRAD generated sufficient accuracy for peak discharge amount and peak time. On the other hands, WRF precipitation data shows the improved lead-time for flood preparation, however peak discharge and peak time were not as good as rain gauge data and NEXRAD. To refine the precipitation accuracy, model parameterization in WRF is necessary to be improved.