Quantitative analysis of social media sensitivity to natural disasters

• Published in: International journal of disaster risk reduction 2019-10, Vol.39, p.101251, Article 101251
• Main Authors: Theja Bhavaraju, Sai Krishna, Beyney, Cyril, Nicholson, Charles
• Format: Article
• Language: English
• Subjects: Natural disasters; Resilience; Social media
• ISSN: 2212-4209; 2212-4209
• DOI: 10.1016/j.ijdrr.2019.101251

As the prevalence of social media real-time communication grows among the public, research has increased regarding its use in various domains of study, including human behavior with respect to natural disasters. Various metrics, whether related to message posting frequency, origination proximity to the disaster, and/or the sentiment of the messages themselves are commonly studied. To the best of our knowledge, no study has been conducted to determine the sensitivity of social media to different types or magnitudes of natural disasters under various circumstances. We select four types of natural disasters (tornadoes, winter storms, wildfires, and floods) and for each we examine multiple recent events along with the associated Twitter behavior to evaluate multiple aspects: duration of social media attention, frequency shifts, frequency shifts for different social vulnerability levels, tweet proximity to the disasters, and sentiment. The results demonstrate that Twitter is indeed a social sensor with different sensitivity levels to natural disasters and depending on the event circumstances, a diverse pattern of social media behavior should be expected. •Sensitivity of Twitter to different types of disruptive events is studied.•Multiple recent disaster events for four types of natural hazards are examined.•Metrics include duration of social media attention, frequency, sentiment, etc.•Results conclusive about different sensitivity levels to different hazards.•Diverse pattern of social media behavior should be expected.