At CIMMS/NSSL, a new multli-radar multi-sensor cloud to ground (CG) lightning probability guidance was created and trained using a random forest machine learning algorithm. As part of the Probabilistic Hazard Information (PHI) Prototype experiment in Hazardous Weather Testbed, an integrated warning team simulated a real-world environment in which forecasters issued PHI and end-users received the output. Forecasters were given automated lightning PHI guidance, which they had the option to modify and/or add a discussion. Upon issuance, emergency managers and broadcast meteorologists received lightning PHI using a web based program called the Enhanced Data Display (EDD). End-users utilized lightning PHI to perform their typical job duties. Research protocols were used to track their decision making and gather feedback.
It is critical that a lightning product can be communicated in an effective manner. The feedback received over the past several years from forecasters and end users is being assessed to improve the visualization and communication of the probabilistic lightning PHI. While forecasters and end-users feel that it is very tedious to watch every lightning object, emergency managers respond more readily to forecaster modified guidance. Lightning poses a particular concern for emergency managers with management responsibilities for outdoor recreation, large venue events, or the transport of sensitive materials or people, yet most lack any lightning information other than recent strikes or simple personal observations Emergency managers consistently voice their approval of lightning PHI, and indicate they would begin using the product right away once available. One of their primary concerns is their ability to receive this information while deployed remotely, and conveying it through rudimentary communication technologies to their partners. While there are obvious uses for emergency managers, the feedback from broadcast meteorologist has been more variable. Broadcasters struggle with the visual appearance of lightning PHI when there are multiple objects. They additionally don’t feel that the hazard often warrants on-air coverage, and the probabilistic thresholds required before they would go on-air are near 100%. Many feel that communication of lightning PHI may primarily be useful on social media platforms.