Retrospective Analysis of the June 18-19, 2024 Oklahoma Panhandle Historic Floods: How Can Probabilistic Data Be Utilized to Improve Impact Decision Support Services (IDSS) and Messaging for Flash Floods?

On the evening of June 18th, the standard forecast for Goodwell, Oklahoma called for a 50% chance of rain through 7:00 AM CDT the next morning. The deterministic Quantitative Precipitation Forecast (QPF) from the Weather Prediction Center (WPC) indicated 0.5” of rainfall through that same time period. With this information alone, most people wouldn’t anticipate waking up to almost 8 inches of rain and historic flooding that night into the next morning. The heaviest rain was expected to fall along a quasi-stationary boundary in Kansas, but the boundary extended further south than expected into an environment which was more primed for heavy rain compared to areas further north. The high QPF potential in the Oklahoma Panhandle was present in the data, but it fell on the extreme end of the distribution of model/ensemble output. The vast majority of model output kept the boundary in Kansas, and forecasters tend to lean towards model consensus which is why the highest WPC QPF amounts of 2+ inches were in Kansas, and not in the Oklahoma Panhandle. The historic flooding triggered multiple ground surveys from the National Weather Service (NWS) and as a part of the flash flood survey project using the Flash Flood Severity Index led by the National Severe Storms Laboratory through which impacts will be cataloged and further researched. Multiple homes were completely flooded, dozens of high water rescues were conducted, and several roads were damaged or completely washed out.

Low-Probability, High-Impact (LPHI) events such as this is where probabilistic forecast information can enhance Impact-Based Decision Support Services (IDSS) and messaging. LPHI events can and do happen, but appropriately messaging these events can be challenging. Tools such as the new Dynamic Ensemble-based Scenarios for IDSS (DESI) make it easier for forecasters to quickly identify critical thresholds within the ensemble data. The utilization of DESI for proper analysis of potential weather threats is critical to successful messaging leading up to impactful events. In this case, 95th percentile 24-hour QPF from the Long Range Ensemble Forecast, National Blend of Models, and High-Resolution Ensemble Forecast showed impactful rainfall amounts extending into the Oklahoma Panhandle, including the town of Goodwell, as early as 24 hours prior to the event onset. Ample data was available via DESI to message a potential flash flood threat should the boundary shift further south compared to the consensus output. This information could have given NWS partners the information they needed to make critical decisions ahead of the floods. Given this, we will present an idealized messaging strategy using probabilistic data ahead of a LPHI event that can be applied across the weather enterprise. We will also show how individual model runs and even consensus output can lead to overconfidence in an outcome, while ensemble data reveals uncertainty. Finally, we will share partner feedback on the messaging and forecast strategy to see how it might have been utilized.