Twitter analytics of three hurricanes in 2017: Harvey, Irma and Maria
Abdinnour, Sue Adeniji, Sesan
Abdinnour, Sue
Adeniji, Sesan
Citations
Altmetric:
Authors
Other Names
Location
Time Period
Advisors
Original Date
Digitization Date
Issue Date
2025-12-17
Type
Article
Genre
Keywords
Heatmaps,Hurricane,Sentiment analysis,Spatial analysis,Topic modelling,Twitter
Subjects (LCSH)
Citation
Abdinnour, S., S. O. Adeniji, and S. Daggubati. 2025. “ Twitter Analytics of Three Hurricanes in 2017: Harvey, Irma and Maria.” Journal of Contingencies and Crisis Management 33: 1–19. https://doi.org/10.1111/1468-5973.70102.
Abstract
This study explores how Twitter functioned as a channel for public communication, emotional expression and situational awareness during the 2017 hurricanes Harvey, Irma and Maria. Using the Twitter API and AWS Elastic MapReduce, a dataset of 203,320 English-language hurricane-related Tweets was collected. The analysis combined a customised AFINN-based lexicon for sentiment evaluation, comparative topic modelling through latent Dirichlet allocation (LDA) and K-Means clustering, and spatial statistical methods to detect geographic clustering. Non-parametric tests, including the Kruskal–Wallis and Mann–Whitney U tests, were used to compare sentiment differences across storms, while chi-square, ANOVA and Kruskal–Wallis analyses examined topic distribution. The findings showed significant differences among hurricanes. Harvey and Irma displayed a clear rebound from negative to positive sentiment after landfall, reflecting hope and early recovery, while Maria remained predominantly negative, signalling extended hardship and dissatisfaction with response efforts. LDA performed better than K-Means, producing coherent seven-topic structures that revealed both shared themes, such as hurricane category, environmental impact, community support and weather updates and unique topics for each event, including presidential response for Harvey, evacuation readiness for Irma and power outages for Maria. Geographical analysis using heatmaps and Moran's I confirmed significant clustering for Harvey and Irma, but diffuse patterns for Maria. Overall, the study demonstrates that integrating sentiment, thematic and spatial analyses provides a multidimensional understanding of disaster communication and recommends that emergency managers employ such frameworks for real-time situational monitoring, geographically targeted relief and adaptive communication strategies to enhance resilience and public trust. © 2025 John Wiley & Sons Ltd.
Table of Contents
Description
Click on the DOI link to access this article at the publishers website (may not be free).
Publisher
John Wiley and Sons Inc
Journal
Journal of Contingencies and Crisis Management
Book Title
Series
Digital Collection
Finding Aid URL
Use and Reproduction
Archival Collection
PubMed ID
ISSN
09660879