Evaluation of short-range ensemble forcasts during the 2003 SPC/NSSL Spring Program

The SPC/NSSL (Storm Prediction Center/National Severe Storms Laboratory) Spring Program is a collaborative experiment held annually in Norman, OK coinciding with the climatological peak of severe convective weather. In 2003, an eight week experiment between 14 April and 6 June evaluated the ability of short-range ensemble forecasts (SREF) to aid in the prediction of severe convective weather. Numerous studies have shown the statistical properties of SREFs are more skillful than a single, higher-resolution deterministic forecast. Yet guidance on applying SREFs to the real-time forecast process are, to a large extent, lacking. As a result, the 2003 Spring Program sought to: (1) determine if SREFs can benefit SPC real-time convective forecasting; (2) explore if SREFs can be integrated into a historically deterministic forecast process; and (3) develop techniques and methodologies which blend the inherently statistical properties of SREFs with expertise provided by operational meteorologists. The experiment consisted of issuing two experimental probabilistic outlooks of severe convection valid for Day 2 (i.e., tomorrow). The first outlook was based on a traditional multi-model approach, while the second outlook incorporated the National Centers for Environmental Prediction (NCEP) SREF. Results based on subjective and objective measures show a small but positive increase in forecast skill when SREF output is included in the forecast process. Several unique diagnostic tools allowing quick interpretation of SREF output are also discussed. These tools range from simple display strategies to more complex diagnostics. An ingredients-based approach in probability space (i.e., combined or joint probabilities) appears very promising, as it combines statistical output from the SREF with situational input provided by the forecaster. Linkages between the higher-resolution deterministic forecast and the SREF were also found to be critically important to the forecast process.