Document Type
Paper
Start Date
16-10-2024 9:50 AM
End Date
16-10-2024 11:10 AM
Abstract
The spread of invasive species requires scalable monitoring techniques to help guide management and conservation strategies. Here, we explore the potential of GeoAI for the rapid validation of invasive species sightings through social media data. We utilized the Flickr API to retrieve 23,000 images from posts that contained the names of 72 invasive species across various taxonomic categories within the contiguous USA. Then using BioCLIP, a computer vision model capable of identifying over 450,000 species, we assessed whether these posts genuinely contained the species that the user tagged. Results varied, with some species like the European Starling showing high match rates, while others like the Alewife had limited matches. This highlights both the strengths and limitations of using GeoAI for ecological monitoring. We also discuss the potential for passive monitoring by analyzing images from specific areas without relying on user-provided species tags. Our findings emphasize the need for scalable, automated solutions to complement manual validation efforts. The study provides reproducible code and instructions via the I-GUIDES platform, enabling ecologists and land managers to implement these tools for real-time monitoring and managing of invasive species. By fostering citizen science initiatives, we aim to enhance the practical application of these methods, ultimately contributing to more effective conservation strategies.
DOI
10.5703/1288284317801
Identifying invasive species sightings from GeoAI-validated social media posts
The spread of invasive species requires scalable monitoring techniques to help guide management and conservation strategies. Here, we explore the potential of GeoAI for the rapid validation of invasive species sightings through social media data. We utilized the Flickr API to retrieve 23,000 images from posts that contained the names of 72 invasive species across various taxonomic categories within the contiguous USA. Then using BioCLIP, a computer vision model capable of identifying over 450,000 species, we assessed whether these posts genuinely contained the species that the user tagged. Results varied, with some species like the European Starling showing high match rates, while others like the Alewife had limited matches. This highlights both the strengths and limitations of using GeoAI for ecological monitoring. We also discuss the potential for passive monitoring by analyzing images from specific areas without relying on user-provided species tags. Our findings emphasize the need for scalable, automated solutions to complement manual validation efforts. The study provides reproducible code and instructions via the I-GUIDES platform, enabling ecologists and land managers to implement these tools for real-time monitoring and managing of invasive species. By fostering citizen science initiatives, we aim to enhance the practical application of these methods, ultimately contributing to more effective conservation strategies.