Next week my co-author Sean Fitzhugh will be presenting our joint work, “Did You Feel It? Spatial Filtering Techniques for Detection of Local Disaster Events”, at the 2014 Annual Meeting of the American Sociological Association. This work comes out of Project HEROIC with colleagues Ben Gibson and Carter Butts at the University of California, Irvine.
Abstract: Spatial filtering is a powerful tool for detecting geographically constrained social phenomena. Filtering allows us to enhance the measurable signal of our phenomenon and ignore known sources of error. In this paper we demonstrate the utility of spatial filtering for detecting disaster events. Such events are typically characterized by disruption of everyday social routines. Those affected by and proximate to these events frequently engage in a variety of response activities: checking the status of family and neighbors, search and rescue, protective action, validation of official information, and message-passing and rumoring behaviors. Informal communication is key component of this response and this increasingly occurs in an online environment. We harness this surge of online, informal communication in response to disaster events by detecting local increases of event-specific keywords. As response to these events is typically geographically constrained (particularly in the immediate aftermath of the event), we use spatial subsampling to increase the resolution of our event-detection techniques. This enables us to identify surges of keyword activity near the event epicenter. By filtering out activity distant to the event, we greatly enhance the signal of event-related activity. We use these spatial filtering techniques to detect a wide variety of disaster events. We consistently find a surge of activity near the site of the event, with the signal tapering off as distance to the epicenter grows. Our results demonstrate the utility of spatial filtering techniques for tuning measurement of hazard-related communication to detect signatures of social processes.