Extended Abstract (260K)P1.18
The development of a new validation technique for tropical cyclone rainfall
Robert Rogers, NOAA/AOML/HRD, Miami, FL; and F. D. Marks, T. P. Marchok, and R. Tuleya
One of the most significant impacts of tropical cyclones is the copious amount of rainfall they often produce. Drowning from inland flooding in landfalling tropical cyclones is the leading cause of death from storms affecting the United States in the past 30 years. For this reason, the Tropical Prediction Center has stated that one of its highest priorities is to improve quantitative precipitation forecasting (QPF) for tropical cyclones. Dynamical numerical models provide one way of forecasting rainfall from tropical cyclones. While such models enable the depiction of the temporal and spatial evolution of tropical cyclones and their associated rain fields, they often exhibit errors related to inadequate initial conditions and model physics. Knowledge of these errors can aid the forecaster in interpreting numerical guidance of rainfall and adjusting their forecasts accordingly.
An accurate diagnosis of rainfall forecast errors requires a validation scheme that accurately measures the performance of the forecast system. However, no standard technique has been developed to validate rainfall forecasts from tropical cyclones. Conventional measures of precipitation forecast skill, such as skill score, are difficult to interpret in the context of tropical cyclones due to the strong dependence of rain location and magnitude on the forecasted track of the storm and differences in the spatial and temporal sampling areas of rain gauge data compared to model output. Therefore, a key task in improving rainfall forecasts is to develop validation schemes for tropical cyclone rainfall that provide a baseline measure of forecast skill independent of track error and sampling issues. In this presentation a new technique will be presented that addresses these issues. The validation scheme will be demonstrated by comparing rainfall forecasts of Hurricane Isabel (2003) from the operational GFDL, GFS, and Eta models and the benchmark Rainfall CLIPER product against observed rain fields provided by the National Precipitation Validation Unit (NPVU) dataset. The validation scheme presented here will enable quantitative comparisons of rainfall from all tropical cyclones covered by the NPVU database (back to 1997).
Poster Session 1, Posters
Wednesday, 5 May 2004, 1:30 PM-1:30 PM, Richelieu Room
Previous paper Next paper