Prior to planning future NWS products and services, it is important that the current accuracy of short-term forecasts provided by the field be assessed. The following questions also require consideration: Do better tools really yield better mesoscale forecast products? How truly beneficial are mesoscale models to this pursuit? Is it advantageous to produce graphical short-term forecast products, versus the traditional text products that have been a staple of the NWS? To answer these questions, NWS Birmingham has undertaken a multi-year operational exercise known as the Summer Convective Rainfall in Alabama Prediction Experiment, or SCRAPE. The primary goals of SCRAPE are to: a) determine a baseline accuracy for warm season short-term probabilistic QPFs in Alabama; b) quantify the overall value of mesoscale model output to the QPF process; and c) assess the relative utility of observed and model-derived fields in this process.
During Summer 2000, NWS meteorologists produced a total of 51 experimental QPF packages for northern Alabama. Each package consisted of five 6-18 hour graphical probabilistic QPFs. The participants had no operational duties on project days, and were able to devote several uninterrupted hours to the preparation of the forecasts. Following an initial evaluation of observed data sources alone, a preliminary “obs-only” 6-hr QPF was generated. Upon completion of this graphic, a thorough assessment of all available model output was made, followed by the creation of the remaining QPF products. In support of this project, the Global Hydrology and Climate Center (GHCC) in Huntsville, AL established an MM5 mesoscale model domain over the Southeast United States with a horizontal grid spacing of 12 km. The output was made available to WFO Birmingham in real-time, and constituted the primary source of model data for the SCRAPE forecasters. Key fields included three-hour accumulated precipitation, low-level equivalent potential temperature and surface temperature.
Preliminary verification results indicate that: a) the SCRAPE forecasts were superior to their operational counterparts, and b) the introduction of high-resolution MM5 output aided in the short-term QPF process. In particular, a comparison of the “obs-only” QPFs with their model-based counterparts yields interesting insights into the relative importance these data types play in the determination of QPF. These results will be presented in detail at the conference.