A major goal of the National Weather Service is to improve community resilience to weather and climate extremes. The verification of extreme quantitative precipitation forecasts (QPF) is critical to improving the prediction of extreme events by identifying and evaluating forecast trends, biases, and errors and by monitoring forecast progress and improvement. A new study benchmarking the performance of extreme QPFs produced by the National Centers for Environmental Prediction’s (NCEP) Weather Prediction Center (WPC) was published in the August 2014 issue of Weather and Forecasting.
In this study, NOAA Hydrometeorology Testbed (HMT) partners from the Physical Sciences Division in the Earth System Research Laboratory (ESRL/PSD) and from the WPC’s Development and Training Branch collaborated to evaluate the performance of the WPC’s 32-km gridded QPFs for extreme precipitation events over an 11 year period (2001-2011). Although WPC is a national center, this analysis calculated and applied regional extreme precipitation thresholds (99th and 99.9th percentile precipitation values) for each River Forecast Center (RFC) region in the continental U.S. (CONUS). WPC QPF performance was evaluated based on five verification metrics: probability of detection (POD), false alarm ratio (FAR), critical success index (CSI), frequency bias, and conditional mean absolute error (MAEcond), which are easily understood by operational forecasters.
Not surprisingly, results show that the extreme events have lower skill in all five verification metrics than lower precipitation thresholds as seen in the figure at right. In addition, skill tends to be lower with longer lead time. However, the verification measures indicate that the 32-km gridded extreme QPFs have incrementally improved in forecast accuracy over the last 11 years, and that the yearly rates of improvement for the extreme events are higher than the improvement rates of the smaller threshold events. In addition, the WPC QPFs tend to verify better in the western and eastern/northeastern U.S. and forecast errors tend to be higher and forecast amounts to be lower in the upper Midwest and south/southwestern U.S.
Contact: Ellen Sukovich
Ellen M. Sukovich, F. Martin Ralph, Faye E. Barthold, David W. Reynolds, and David R. Novak (August 2014): Extreme quantitative precipitation forecast performance at the weather prediction center from 2001 to 2011. Wea. Forecasting, 29, 894-911. doi: 10.1175/WAF-D-13-00061.1