Broadcast Meteorologist Use of Social Media in the 2018 Hazardous Weather Testbed Probabilistic Hazard Information Project

Hazardous weather information is constantly being improved upon by NOAA to empower individuals as they decide how to respond during severe weather. The NOAA Hazardous Weather Testbed (HWT) is a conceptual framework and physical space where researchers, National Weather Service forecasters, and users of hazardous weather information participate in experiments to improve hazardous weather information. This particular HWT experiment focuses on the use of probabilistic hazard information (PHI) to better communicate severe weather impacts. Since broadcast meteorologists are the main communicators of severe weather information to the public, it is important to study this group and assure that they can access, understand, use and transmit probabilistic information effectively. For three weeks, two broadcast meteorologists simulated television and social media coverage using PHI during past severe weather cases involving hail, wind, lightning and/or tornadoes. Coverage decisions included running crawls, posting to social media, and cutting into TV programming when necessary - all while using the new probabilistic information. This presentation explores how broadcast meteorologists’ social media use changed with the use of PHI. Researchers accessed real-world tweets from the displaced real-time cases simulated during the project, and used these tweets as a baseline to compare the tweets made during this experiment with PHI. The participants’ current social media pages were also researched in order to compare their typical social media coverage. Results showed that probabilistic information changed how broadcast meteorologists portrayed specific threats in many aspects including geography, time of arrival, instructional information, the use of legacy polygon warnings, impacts mentioned, and storm motion. The main difference shows how specificity in most every aspect was used more frequently in the cases during this experiment compared to the past actual events due to probabilistic information being provided. Additional results will be shared.