Evaluating elements of risk communication in risk messaging from officials on Twitter in the 2021 Marshall fire

Abstract

The Marshall re in Boulder County, Colorado sparked and spread into nearby urban areas in an unusually fast timeframe. Without a widely used warning system to alert the public, o cials used social media as one warning information outlet to encourage rapid, large-scale evacuations. Tweets (N = 1,338) published by 109 media and government Twitter accounts on the day of the re were examined for risk messaging components and coded according to which components were included. The risk messaging components used were derived from the Protective Action Decision Model and the Extended Parallel Process Model. Tweets from the National Weather Service in Boulder were evaluated using recommendations from NOAA's Hazard Risk Communication guide. Some types of environmental cues improved several engagement metrics for tweets from media, and emergency managers and rst responders. Social cues only effected engagement for local media accounts, though there were mixed results depending on the variables included in the regression model. Inclusion of actionable information boosted engagement for government o cials, though this result also depended on which variables were included in the model. Inclusion of threat information improved some or all engagement metrics for journalists, local newspapers and non-traditional news websites, government, and emergency management accounts. The National Weather Service ful lled the eleven recommendations from NOAA's HRC guide that were used in this research, and the use of lively language, vivid images, empathy, and the "all-clear" were noted among the well-performing tweets. This study provides helpful information to those posting on social media in an o cial capacity during a wild re by examining how risk communication theory and recommendations performed in a real-world scenario.