Translating Social Media Crisis Narratives into Road Network Utilization Metrics: The Case of 2020 Oklahoma Ice Storm

TL;DR

This study analyzes how social media crisis narratives during the 2020 Oklahoma ice storm influenced road network utilization, combining social media data, natural language processing, and network analysis to understand infrastructure impacts.

Contribution

It introduces a novel methodology integrating social media analysis with geospatial and network techniques to assess infrastructure response during a disaster.

Findings

Social media narratives correlate with traffic disruptions.

Geospatial analysis reveals specific network vulnerabilities.

Insights inform disaster communication and infrastructure resilience.

Abstract

Risk communication in times of disasters is complex, involving rapid and diverse communication in social networks (i.e., public and/or private agencies; local residents) as well as limited mobilization capacity and operational constraints of physical infrastructure networks. Despite a growing literature on infrastructure interdependencies and co-dependent social-physical systems, an in-depth understanding of how risk communication in online social networks weighs into physical infrastructure networks during a major disaster remains limited, let alone in compounding risk events. This study analyzes large-scale datasets of crisis mobility and activity-related social interactions and concerns available through social media (Twitter) for communities that were impacted by an ice storm (Oct. 2020) in Oklahoma. Compounded by the COVID-19 pandemic, Oklahoma residents faced this historic ice storm (Oct. 26, 2020-Oct. 29, 2020) that caused devastating traffic impacts (among others) due to excessive ice accumulation. By using the recently released academic Application Programming Interface (API) by Twitter that provides complete and unbiased data, geotagged tweets (approx. 210K) were collected covering the entire state of Oklahoma, and ice storm-related tweets (approx. 14.2K) were considered. First, the study uses natural language processing and text quantification techniques to translate crisis narratives (i.e., tweets). Next, geo-tagged tweets are mapped into co-located road networks using traditional GIS techniques. Finally, insights are generated using network science theories and quantified social narratives to interpret different elements of road networks (e.g., local roads, freeways, etc.) for the Oklahoma communities that were impacted by the ice storm event during the pandemic.