On January 3, 2018, a significant winter storm struck northeast Florida and southeast Georgia, an event that occurs once every 25-50 years…the last one occurred in 1989.  A challenging forecast faced the staff at WFO Jacksonville, as some of the models were forecasting the event to occur. As this hasn’t happened in 30 years, there was no knowledge on model accuracy or bias in predicting such a rare event in Florida. In the past, models were notoriously bad in forecasting shallow-cold-air events.

While the global models had problems predicting the event, higher-resolution options accurately predicted precipitation type and type-changeover.  Despite the known NAM bias of over-predicting precipitation in cold air (it consistently showed 8-12 inches of snow in southeast Georgia), its precipitation type forecasts were accurate. As the event drew closer, the HRRR (and parallel HRRRx out to 36 hours) precisely forecast both precipitation type and amount. 

This supported accurate winter storm watches/warnings, and decisions not to include Jacksonville and Gainesville in them, despite the fear of past model handling of shallow cold air. The challenge to forecasters was whether or not to believe the higher-res model forecasts of such a rare event, particularly one coming out of the Gulf of Mexico, a warm air source that makes winter storm events so rare.

The accurate watches/warnings issued in advance, and DSS webinars for nearly 100 partners that focused as much on impact as weather, supported efforts to keep people off the roads. During state calls, Florida Highway Patrol officials noted the lack of accidents and injuries/fatalities on icy roads.

As observed by SOO Pete Wolf, “the improvements in higher-resolution modeling were key to accurately warn of a high-impact event that probably would have been missed 10-15 years ago.”