The April tornado outbreak that moved through the central Plains and lower Mississippi Valley last month produced 34 confirmed tornadoes across three days, caused four fatalities, and generated several hundred million dollars in property damage. It also produced a significant forecast failure: the most severe day of the outbreak was not the day the models had highlighted as the highest-risk period, a gap that the Storm Prediction Center has been analysing in some detail since.

What the models missed

The outbreak's highest-impact day developed along a secondary dryline boundary that the operational models had assigned low probability of being the focus of convective initiation. The primary boundary — the one that received the high-risk designation — produced significant thunderstorms but fewer tornadoes than forecast. The secondary boundary, which received a much lower categorical designation, was where the most significant supercells formed.

In retrospect, the models handled the large-scale atmospheric setup reasonably well but underperformed on the mesoscale features that determined which boundary would be the focus of the most significant convection. The distinction is important because the categorical risk designations — which drive public response — are based on the model ensemble guidance, which did not distinguish between the two boundaries until too late to revise the public messaging.

The warning lead time question

The warning lead time on the individual tornadoes that caused the fatalities ranged from six to fourteen minutes — within the normal operational range but below the longer leads that had characterised some of the more successful recent outbreaks. The National Weather Service offices covering the affected area have flagged the lead time distribution as something they intend to examine in the post-event survey.

What is changing

The Storm Prediction Center is piloting a revised approach to outlooks that provides more explicit uncertainty framing around which boundary will be the focus of the highest-risk convection when multiple candidates exist. The revision does not change the categorical designations but adds probabilistic language to the outlook discussions that should allow experienced emergency managers to understand the forecast uncertainty better than the categorical designations alone convey.