Cover: Lessons from Harvey

Lessons from Harvey

Improving Traditional Damage Estimates with Social Media Sourced Damage Estimates

Published in: Cities (2021). doi: 10.1016/j.cities.2021.103500

Posted on Nov 16, 2021

by Carlos Villegas, Matthew J. Martinez

Social media systems and crowdsourced data sites were incredibly active during disasters. Residents, first responders, and officials all turn to these systems to impart information and make calls for assistance. These systems will likely continue to hold a central informational and communication role in future disasters. Analyzing the trends and information that come from these sources in real-time aids the recovery process and help public agencies, first responders and researchers more quickly assess damages during and immediately after a disaster. Traditional sources, such as the initial FEMA damage estimates can miss areas of heavy impact and are often time delayed by several weeks. Using Harris County, Texas in the Houston region and the 2017 Hurricane Harvey flooding, the study provides a novel use-case in crisis informatics. The study leverages calls for help during the flooding event to mine address-level information to proxy damage estimates at the parcel-level. The study finds 36- to 53% of Twitter-sourced damage estimates are not captured in the FEMA estimates, significantly augmenting initial estimates with new data—feasibly within hours after the information is first tweeted. Empirically, the study evaluates how parcel-level FEMA damage estimates and Twitter-sourced damage estimates complement each other to proxy damaged structures.

Research conducted by

This report is part of the RAND external publication series. Many RAND studies are published in peer-reviewed scholarly journals, as chapters in commercial books, or as documents published by other organizations.

RAND is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.