Using social media to detect and locate wildfires
Methods for detecting and tracking natural hazards continue to increase in coverage, resolution and reliability. However, information on the social impacts of natural hazards is often lacking. Here we test the feasibility of using social media data (Twitter and Instagram) to detect and map an important class of natural hazard: wildfires. We analyse social media posts associated with wildfires over several time periods and compare them with wildfire occurrence data derived from satellite-based remote sensing data and on-the-ground observations. For the whole of the contiguous United States, we find significant temporal correlations between wildfire-related social media activity and wildfire occurrence, but also that there is substantial variation in the strength of this relationship at smaller spatial scales (states and counties). We then explore the utility of social media for location of wildfire events, finding good evidence to support further development of such methods. We conclude by discussing several challenges and opportunities for application of this novel data resource to provide information on impacts of natural hazards.
The authors were supported by a Research on Changes of Variability and Environmental Risk (RECoVER) grant funded by EPSRC (EP/M008495/1).
This is the final version of the article. Available from AAAI via the URL in this record.
Technical Report WS-16-20. 10th International AAAI Conference on Web and Social Media (ICWSM 16), W8: Social Web for Environmental and Ecological Monitoring, 17-20 May 2016, Cologne, Germany