The topics discussed and developed in this dissertation strive to propose a conceptual model describing how information sources and content characteristics affect social-mediated disaster communication outcomes in the form of communication strategies, as well as audience perceptions and behavioral engagement. The model proposes that the role that information sources play in social media content creation before, during and after disasters is shaped by broader routines, societal and systemic factors. It also contends that content characteristics affect the way in which audiences engage with messages on social media. The major tenets of the model were tested through a content analysis of disaster-related tweets posted before, during and after Hurricane Matthew, and a 2 (Post Modality: Image- vs. GIF-based Social Media Disaster Preparedness Posts) x 3 (Visual Focus: Reactive- vs. Proactive- vs. Hero-themes Social Media Post Visuals), also including an additional text-based condition, between-subjects online experiment. Content analysis results found that social media users were more likely to retweet, like and reply to image-based posts rather than text-based posts. The same was true for users engaging with video-based posts rather than image-based posts. In turn, the online experiment results found that affective risk perception plays a mediation role in the relationship between previous hurricane experience and three target communication outcomes: crisis information sharing intentions, crisis information seeking intentions, and guidance adoption intention. The practical and theoretical implications of these findings are discussed.