Date of Award

Summer 8-2017

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Geography and Geology

Committee Chair

Bandana Kar

Committee Chair Department

Geography and Geology

Committee Member 2

Chaoyang Zhang

Committee Member 2 Department

Computing

Committee Member 3

David M. Cochran

Committee Member 3 Department

Geography and Geology

Committee Member 4

Carl A. Reese

Committee Member 4 Department

Geography and Geology

Committee Member 5

George T. Raber

Committee Member 5 Department

Geography and Geology

Abstract

While Twitter has been touted to provide up-to-date information about hazard events, the relevance and reliability of tweets is yet to be tested. This research examined the relevance and reliability of risk information extracted from Twitter during the 2013 Colorado floods using five different approaches. The first approach examined the relationship between tweet volume and precipitation amount. The second approach explored the relationship between geo-tagged tweets and degree of damage. In the third approach, the spatiotemporal distribution of tweets was compared with flood extent. In the fourth approach, risk information from tweets were compared with survey responses obtained in a Department of Homeland Security report about risk communication to determine what people expect to be included in alerts vs. what is communicated via tweets. In the fifth approach, tweets containing top frequent keywords and hashtags were compared with official reports using cosine similarity method. For reliability assessment, contents of relevant tweets were manually compared with official data and images.

The findings indicated that relevant tweets provided information about the event, its impacts, and contained other risk information that public expects to receive via alert messages. Content analysis of images revealed that tweets were also reliable in disseminating information about damages and impacts. Given that the Crowdsourcing and Citizen Science Act (2016) authorizes agencies to use crowdsourcing to increase public response to emergency alerts, the methodology used in this study could be used by emergency management personnel (EMP). The findings could also be used by EMPs to identify relevant and reliable tweets. However, out of 1 million English tweets, 14% were relevant, 3% were reliable, and 0.44% were geo-tagged to Boulder. Although the geographically relevant tweets could have eliminated possible misinformation shared by “outsiders”, very limited percentage of social media was useful, relevant, and reliable. Furthermore, social media analytics was time consuming and computationally intensive, which may not be feasible for EMP. Future research should focus on developing a matrix to assess data quality of crowdsourced data, automating implementation of analytics, and developing a citizen-science based approach to gather focused data about hazard events.

ORCID ID

0000-0002-4161-0388

Share

COinS