Representation and conceptual drift in social vulnerability models: the case of Hispanic representation in social media and FEMA data of Hurricane Harvey

Authors: Joseph Holler*, Middlebury College, Claire Gomba*,
Topics: Hazards and Vulnerability, Geographic Information Science and Systems
Keywords: Hazards, social vulnerability, social media
Session Type: Paper
Day: 4/5/2019
Start / End Time: 8:00 AM / 9:40 AM
Room: Forum Room, Omni, West
Presentation File: No File Uploaded


Social vulnerability describes the predisposition of exposure units (people, households, neighborhoods, etc.) to experience harm from natural hazards, and can be modelled with geographic information systems by aggregating its component indicators into an index. Indicators for deductive models are selected with theory or expert/stakeholder opinion; while indicators for inductive models are selected for statistically significant relationships with harmful outcomes. For both modelling approaches, validation requires an indicator of harmful outcomes, for which social media posts of distress during a disaster may prove valuable. Geographic social media data is increasingly available during disasters for the purpose of seeking assistance and coordinating rescue and recovery work. During the 2017 Hurricane Harvey disaster, thousands of social media users posted messages of distress while volunteers transformed social media data into a spreadsheet and dynamic map. The social media activity was publicized by local news outlets during the disaster and used to guide rescue efforts. Previous research sought to validate prominent social vulnerability indices using Hurricane Harvey social media distress data, revealing a significant negative relationship between Hispanic populations and harmful outcomes. This research uses models of flooding, records of flood insurance and disaster assistance claims, and historical twitter data to investigate whether risk and harm in Hispanic neighborhoods is underrepresented in social media and FEMA databases. Unrepresentative indicators of harm should not be used for model validation or inductive model development, or social vulnerability models risk conceptual drift and misdirection of resources for disaster risk reduction, response, and recovery.

Abstract Information

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

To access contact information login