Detecting Local Crisis Events: A Case Study on the Colorado Wildfires through Social Media and Satellite Imagery

Social media has become vital for disseminating natural disaster information, but its reliance has limitations. This study underscores the crucial integration of diverse data sources, particularly remote sensing and social sensing, to overcome challenges faced by traditional uni-modal methods. Using the Colorado wildfire as a case study, we applied advanced techniques, including named entity recognition and natural language processing, to extract tweet locations, grouping them based on geographical relevance. To enrich satellite imagery, we extracted images from tweet URLs, associating them with relevant tweet groups. This research combines theoretical and practical insights, addressing detection challenges within a simplified framework. It highlights the innovative integration of social media and satellite imagery, offering valuable directions for disaster response and management. This research aims to make the linkage between social media data and remote sensing data more clear to researchers. Specifically, it aims to enrich satellite imagery with images and text from tweets to provide a clearer view of events and their locations.