Twitter as a Potential Disaster Risk Reduction Tool. Part III: Evaluating Variables that Promoted Regional Twitter Use for At-risk Populations During the 2013 Hattiesburg F4 Tornado

Abstract

INTRODUCTION: Study goals attempt to identify the variables most commonly associated with successful tweeted messages and determine which variables have the most influence in promoting exponential dissemination of information (viral spreading of the message) and trending (becoming popular) in the given disaster affected region. METHODS: Part II describes the detailed extraction and triangulation filtration methodological approach to acquiring twitter data for the 2013 Hattiesburg Tornado. The data was then divided into two 48 hour windows before and after the tornado impact with a 2 hour pre-tornado buffer to capture tweets just prior to impact. Criteria-based analysis was completed for Tweets and users. The top 100 pre-Tornado and post-Tornado retweeted users were compared to establish the variability among the top retweeted users during the 4 day span. RESULTS: Pre-Tornado variables that were correlated to higher retweeted rates include total user tweets (0.324), and total times message retweeted (0.530). Post-Tornado variables that were correlated to higher retweeted rates include total hashtags in a retweet (0.538) and hashtags #Tornado (0.378) and #Hattiesburg (0.254). Overall hashtags usage significantly increased during the storm. Pre-storm there were 5,763 tweets with a hashtag and post-storm there was 13,598 using hashtags. CONCLUSIONS: Twitter's unique features allow it to be considered a unique social media tool applicable for emergency managers and public health officials for rapid and accurate two way communication. Additionally, understanding how variables can be properly manipulated plays a key role in understanding how to use this social media platform for effective, accurate, and rapid mass information communication.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 9 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.