Evaluation of Quantitative Precipitation Forecasts from the GFDL Hurricane Model

Over the past several decades considerable progress has been made in the ability to forecast the tracks of tropical cyclones. Some modest intensity forecast improvement has also been obtained. Since 1970, however, the largest loss of life in the U.S. from landfalling tropical cyclones has resulted from inland flooding. The operational version of the Geophysical Fluid Dynamics Laboratory (GFDL) hurricane model provides forecasts of track, intensity, structure and rainfall. A comprehesive verification of the track and intensity forecasts is routinely performed by the National Hurricane Center, but the structure and rainfall forecasts have not been carefully evaluated. In this paper, the rainfall forecasts from the GFDL model will be evaluated by comparison with rain gauge data. The observed post-landfall rainfall rate in a storm-relative coordinate system determined from a historical climatology (1946-1999) will first be described. The observed rainfall rate will be compared to that from the GFDL model for the period 1995-1999. Individual forecasts from the model will also be compared to gauge data. The contributions to the rainfall forecast errors from track errors and from the rain rate errors will be estimated.