Using Twitter to Analyze the Effect of Hurricanes on Human Mobility Patterns

Abstract

Understanding human mobility patterns becomes essential in crisis management and response. This study analyzes the effect of two hurricanes in the United States on human mobility patterns, more specifically on trip distance (displacement), radius of gyration, and mean square displacement, using Twitter data. The study examines three geographical regions which include urbanized areas (Houston, Texas; Miami-Dade County, Florida) and both rural and urbanized areas (North and South Carolina) affected by hurricanes Matthew (2016) and Harvey (2017). Comparison of movement patterns before, during, and after each hurricane shows that displacement and activity space decreased during the events in the regions. Part of this decline can be potentially tied to observed lower tweet numbers around supply facilities during hurricanes, when many of them are closed, as well as to numerous flooded and blocked roads reported in the affected regions. Furthermore, it is shown that displacement patterns can be modeled through a truncated power-law before, during, and after the analyzed hurricanes, which demonstrates the resilience of human mobility behavior in this regard. Analysis of hashtag use in the three study areas indicates that Twitter contributors post about the events primarily during the hurricane landfall and to some extent also during hurricane preparation. This increase in hurricane-related Twitter topics and decrease in activity space provides a tie between changed travel behavior in affected areas and user perception of hurricanes in the Twitter community. Overall, this study adds to the body of knowledge that connects human mobility to natural crises at the local level. It suggests that governmental and rescue operations need to respond to and be prepared for reduced mobility of residents in affected regions during natural crisis events. disasters including wildfires, earthquakes, floods, storms, tsunamis, and hurricanes [6,7]. As expected, the frequency of disaster related tweets is highest in the spatial proximity of a disaster [8,9].Related work on movement analysis focuses primarily on the effects of crisis events on a larger scale and longer-term movements. One study, for example, explored evacuation patterns by leveraging user location information from tweets posted in the hours prior and concurrent to Hurricane Matthew in 2016 [10], and another study used Twitter data to estimate the percentage of evacuees during the same hurricane [11]. Tweets were also used to identify refugee migration patterns from the Middle East and Northern Africa to Europe during the initial surge of refugees aiming for Europe in 2015 [12].Despite such analysis, the effect of natural disasters on local mobility patterns (e.g., for the population remaining in the affected regions during such an event), is less explored. Furthermore, the implications of changed mobility patterns on planning and rescue operations need to be addressed. As a step in this direction, the goal of this study is to investigate changes in human activity patterns du...

An analysis of twitter as a relevant human mobility proxy

“…For example, the work in [6] focuses on forecasting dengue disease transmission by studying the public transport data, related Twitter posts and the incidence of such a disease in tropical areas. Similarly, the work in [1] aims to detect natural disaster situations by means of TWT posts without the need for geo-tagged data. The authors rely instead on the use of other features such as posting time, publication frequency and hashtag trends.…”

“…To achieve this goal, we propose a comprehensive comparison between an official study about the human mobility in Spain developed by the Spanish National Institute of Statistics (INE) 1 , which is based on a nation-wide mobile-phone location dataset, with a social media dataset crawled from TWT covering the same spatial and temporal dimensions. More in detail, both datasets cover a 3-month period from April to June, 2020.…”

Quality Characteristics for User-Generated Content

“…g is the data value, and N is the number of rules for g (4) [59]. (4) Credibility: How data are accepted or regarded as true, real, and credible, where dist is the distance between the sensor s and entity e, and dmax is the maximum distance acceptable (5) [60]. (5) Objectivity: Data is unbiased and impartial, where E is evidence, H is a hypothesis (assumed value), and p() denotes the probability (6) [61].…”

“…Social media platforms such as Facebook, Twitter, and Instagram, crowdsourcing platforms such as Wikipedia and OpenStreetMap, and citizen science platforms such as eBird and iNaturalist, are examples of UGC gathering platforms. UGC has been demonstrated to be used for investigating customer feedback [2,3], monitoring catastrophic environmental effects [4], tracking visitors in protected areas [5], flood research [6], emergency reporting [7], future prediction [8], service quality analysis [9], managing online encyclopedia [10], and targeting advertisements and recommendations for potential customers [11,12].…”

Understanding human mobility using Twitter

“…In another related study [46], the authors investigate the effect of hurricanes on human movement patterns in the United States using Twitter data. They check for changes in trip distances, radii of gyrations and mean-squared displacements, and compare these metrics before, during and after each hurricane.…”