Temporal patterns and life cycle dynamics of social media user activity during disasters: A data-driven approach for effective crisis communication

Situational awareness of social media user activity during disasters is essential for effective and efficient crisis communication. Social media platforms such as Twitter and Facebook have emerged as critical channels for governments, relief agencies, and volunteer organizations to disseminate important information to the public during disasters. However, existing models are often limited to a specific disaster event, lacking a comprehensive and generalized approach. This paper introduces a novel life cycle concept to investigate the dynamics of social media engagement during multiple wildfires and bombing/shooting events in the United States. We extensively analyze millions of tweets and identify significant temporal commonalities, unveiling generic diurnal and life cycle patterns within each disaster category. This study proposes two models for the Hourly Distribution of Tweets and Daily Distribution of Tweets to explore different evolutionary stages characterizing diurnal and life cycle patterns. The hourly model accurately calculates peak engagement times for wildfire and bombing/shooting events, with peak engagement occurring between 9:00 AM and 12:00 PM. The daily models reveal that the highest diffusion rate is observed between two to four days from the wildfire formation day, while the peak engagement of a bombing/shooting event comes within the first 24 h of the incident. The findings enhance the understanding of social media dynamics during disasters and assist crisis communication agencies in developing communication strategies that ensure prompt and effective dissemination of critical information to affected populations.