Towards Intelligent Disaster Management: A Spatial-Temporal Analysis of Rescue Requests on Twitter during Hurricane Harvey

Authors: Danqing Liao*, Texas A&M University, Lei Zou, Texas A&M University, Michelle Meyer, Texas A&M Univeristy, Nina Lam, Louisiana State University, Dongying Li, Texas A&M University
Topics: Hazards, Risks, and Disasters, Spatial Analysis & Modeling
Keywords: Social Media, Hurricane Harvey, Online Rescue, Geographic Information Retrieval, Disaster management
Session Type: Paper
Day: 4/9/2020
Start / End Time: 9:35 AM / 10:50 AM
Room: Governors Square 12, Sheraton, Concourse Level
Presentation File: No File Uploaded


Although evacuation warnings have been issued before Hurricane Harvey's landfall, many Houston residents failed to evacuate in time. As rain barreled down on Houston through the weekend, 911 emergency lines were overwhelmed with more than 56,000 calls in the Houston area on August 26-27, 2017. As a result, many residents resorted to social media to call for rescue. This changing use of social media marks Harvey as one of the very first disastrous events in which social media played an important role in fast-responding rescue missions. To improve the efficiency of social media use for rescue in future events, it is important to analyze the behaviors of online help-seekers during disasters. This study analyzed the spatial-temporal patterns of rescue requests on Twitter during Hurricane Harvey from Aug 25 to 31, 2017. The objectives are three-fold: (1) to develop a rescue request detection algorithm to identify tweets requesting for assistance; (2) to develop a geographic information retrieval algorithm to extract location information from tweet contents; (3) to investigate the spatial-temporal patterns of online rescue requests. Each tweet is classified as requesting for rescue or not by natural language processing combined with deep neural networks. For tweets requesting for rescue, we extract and geocode the address information in the tweet contents to analyze the spatial-temporal patterns and socio-environmental characteristics of online help-seekers. The developed geodatabase and algorithms will guide evacuation preparedness for future disastrous events at Houston and shed light on building artificially intelligent systems for disaster management.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login