Forecasting the Super Tuesday tornado outbreak at the Storm Prediction Center: Why forecast uncertainty does not necessarily decrease as you get closer to a high impact weather event

The 5 February 2008 tornado outbreak ranks as the deadliest tornado outbreak in 23 years for the United States, and the deadliest since 3-4 April 1974 in Kentucky and Tennessee. Eighty-four tornadoes occurred during the course of the outbreak, killing 56 people directly. Synoptically, the potential for an outbreak of severe thunderstorms and tornadoes was evident up to a week prior to the outbreak; SPC severe weather outlooks began highlighting the affected region beginning six days in advance on 31 January. Subsequent SPC forecast products not only provided increasingly greater spatial details of the severe weather threat, but also heightened the potential significance of the outbreak in the days leading up to the event. The SPC severe weather outlook products culminated in an upgrade to ‘moderate risk' on the Day 2 Outlook issued at 06 UTC 4 February and a final upgrade to ‘high risk' on the Day 1 Outlook issued at 13 UTC 5 February.

Although it is commonly thought that forecast uncertainty should decrease as the lead time to a significant weather event diminishes, this is often not the case for severe thunderstorm and tornado forecasting. This results from the more limited predictability of both the mesoscale pre-convective and near-storm environments, which provide a controlling influence on convective development and evolution, and the individual thunderstorms themselves, which generally have low predictability. Thus, it is not uncommon that key details such as timing of thunderstorm initiation, mode, and intensity are often not known even several hours in advance. In fact, for the 5 February outbreak the ‘ramping up' of the severe risk as the event neared was particularly challenging, as the likely details of thunderstorm development became murkier 12-18 hours in advance as additional observational and numerical model data became available. These data increased concerns that the cap would hold ahead of an approaching surface cold front, and storms associated with a subtle, southern stream mid-upper level impulse might overspread much of southern Missouri, northern Arkansas and western parts of Kentucky and Tennessee during the day stabilizing a large portion of the region. The difficult operational challenges in assessing increasing volumes of higher resolution data in the forecast process, especially when these data conflict with other model data, observations and conceptual models of the atmosphere, will be illustrated by this case.