Verification of probabilistic severe storm forecasts at the SPC

Since March 1999 the Storm Prediction Center (SPC) has issued twice daily experimental probabilistic severe weather forecasts. These forecasts currently augment the traditional Convective Outlooks. The Convective Outlook is the primary forecast used by a wide variety of users such as National Weather Service offices, emergency managers, and other organizations for national-scale forecasts of severe weather on time scales from several days down to several hours. The Convective Outlook is a categorical forecast of severe weather which depicts the threat of severe weather in terms of "risk" (slight, moderate, high). This structure does not explicitly state the forecaster's expectations of the threats of individual severe weather hazards (hail, wind, and tornadoes) such that the user cannot easily make decisions based on the forecast. The SPC is attempting to address these problems by producing subjective probabilistic forecasts of individual severe weather hazards. Probabilistic forecasts directly express uncertainty unlike categorical forecasts where such information is hidden in imprecise wording. Probabilistic forecasts of the individual hazards provides the information directly to the user who can then make their own decisions rather than first having to attempt to weigh the uncertainty implied in categorical statements before choosing to act.

Probabilistic forecasting of rare events presents many challenges both in terms of design of the forecasts themselves and the development of a meaningful verification system. Several important aspects of the system will be discussed including the development of a basic climatology of severe weather, the choice of appropriate probabilities for the various threats, and a framework for transforming daily severe weather reports into probabilities. Initial results are very encouraging and suggest that the SPC forecasters perform quite well at such forecasts.

An important aspect of the verification of any forecast is a numerical measure of the skillfulness embodied within it. Initial attempts to assess skill scores for the SPC probabilistic forecasts have proven difficult to interpret. Skill is typically assessed relative to a baseline forecast (usually climatology). The daily climatological threat of the individual hazards is typicall less than 5%. Such small values of climatology leave little room for skillful forecasts unlike probability of precipitation forecasts which benefit from typical values of climatology that are near 20% in many locations.