Introduction
During natural disasters, social media platforms like Twitter transform into a "digital firehose" of real-time information. While emergency responders increasingly utilize this data for immediate situational awareness, a significant challenge remains: integrating this chaotic, high-volume data into formal disaster management processes. This research addresses the gap in understanding how communication evolves throughout the entire disaster lifecycle, moving beyond the immediate response phase to include long-term recovery.
Methodology
The study employed a hybrid approach combining qualitative social science techniques with computational data processing. Researchers analyzed a dataset of approximately 8.8 million tweets from the 2013/14 UK storms and floods. Using Python-based filtering, they extracted a representative sample of 638 tweets for manual thematic coding.
Central to the methodology was mapping these communications to Faulkner’s Disaster Management Lifecycle (DML), which categorizes events into six phases: Pre-Event, Prodromal (imminent), Emergency, Intermediate, Long Term, and Resolution. This allowed the authors to observe how public discourse shifts as a disaster progresses from an imminent threat to an emergency and eventually into recovery.
Thematic Framework
The researchers identified 25 "Basic Themes" which were aggregated into three "Organised Themes" to capture the nature of the discourse:
- Subjective and Critical: Emotional or urgent statements including warnings, sympathy, and complaints.
- Informative and Supportive: Practical information shared with a helpful tone, such as preparation tips.
- Objective and Instructive: Factual, logistical data like school closures or transport updates.
Results and Discussion
The analysis revealed a distinct shift in the type of content shared as the crisis evolved. In December 2013, "Subjective and Critical" tweets were the most prevalent. However, as the flooding persisted into January and February 2014, "Informative and Supportive" and "Objective and Instructive" categories grew in proportion, reflecting a shift toward community resilience and logistical management.
A key finding was the cyclical nature of the disaster. The 2013/14 floods did not follow a linear path through the DML; instead, the crisis frequently "reset" from recovery phases back to imminent threat phases as successive storms hit. Twitter activity mirrored these cycles, with volume peaking in February 2014, which accounted for 65% of the analyzed data.
Conclusion
This research demonstrates that social media provides a nuanced record of public sentiment and needs that extends beyond tactical alerts. While objective data is crucial for immediate response, "Subjective" tweets offer invaluable insights for the recovery phase by highlighting public dissatisfaction and specific pain points. Future work aims to use these manually identified themes to train machine learning classifiers, enabling disaster managers to monitor public sentiment in real-time and improve long-term resilience policies.
Summary updated