Document Type
Paper Presentation
Start Date
6-10-2023 2:30 PM
End Date
6-10-2023 3:15 PM
Abstract
While social media data are increasingly being used in the study of pressing environmental problems, their ability to monitor environmental changes has scarcely been assessed. Understanding this viability is highly important as climate change increasingly impacts public health, and behavior. We examine social media photographs associated with wildfires in Yellowstone National Park to assess if images and content can adequately capture environmental change associated with large-scale landscape impacts - wildfires - using computer vision, natural language processing and spatiotemporal analysis. We find that social media posts associated with wildfire events rarely capture the fires themselves, while landscape impacts including burnt trees and early succession are more frequently the topic of photography. Furthermore, we find that computer vision has challenges with capturing these phenomena. While capturing wildfires proved difficult, developing multimodal analysis including natural language processing, spatial, trend and computer vision analysis at scale may open opportunities for more general understanding of social media’s efficacy for monitoring environmental change.
DOI
10.5703/1288284317679
Can Social Media Help Us Understand The Impact of Climate Change on Forests in The US?
While social media data are increasingly being used in the study of pressing environmental problems, their ability to monitor environmental changes has scarcely been assessed. Understanding this viability is highly important as climate change increasingly impacts public health, and behavior. We examine social media photographs associated with wildfires in Yellowstone National Park to assess if images and content can adequately capture environmental change associated with large-scale landscape impacts - wildfires - using computer vision, natural language processing and spatiotemporal analysis. We find that social media posts associated with wildfire events rarely capture the fires themselves, while landscape impacts including burnt trees and early succession are more frequently the topic of photography. Furthermore, we find that computer vision has challenges with capturing these phenomena. While capturing wildfires proved difficult, developing multimodal analysis including natural language processing, spatial, trend and computer vision analysis at scale may open opportunities for more general understanding of social media’s efficacy for monitoring environmental change.
Comments
I am adding the Pdf of this manuscript per the organizers comments (added the latex files previously)