Embracing human noise as resilience indicator: twitter as power grid correlate

Abstract

There are typically two approaches for measuring disaster resilience: technically dynamic measures produced by sensors attached to physical objects and socially static metrics that engage demographic indicators within a given geographic location. Although these approaches allow resilience to be represented before and after disruption, it can be difficult to measure resilient behavior during an event. We propose that social media data can be used to nowcast the ongoing state of critical infrastructure during a disaster. Through an analysis of tweets made during Hurricane Sandy and power outage data obtained after the event, we find that tweets that mention power, utility, or electricity were correlated with loss of power. We conclude with a discussion of barriers to realizing this concept.