Data Skeptic Bird Distribution Modeling with Satbird
Sep 10, 2024
Mélisande Teng, a PhD candidate at Université de Montréal, dives into her groundbreaking research on biodiversity monitoring using remote sensing and computer vision. She discusses the innovative Satbird project, which enhances bird distribution modeling by combining satellite data and citizen science. The conversation highlights challenges like data imbalance in different regions and the importance of acoustic monitoring. Mélisande also explores the intricacies of joint species distribution modeling and advocates for collaboration between machine learning and ecology to advance conservation efforts.
AI Snips
Chapters
Transcript
Episode notes
Power of Satellite Data and Embeddings
- Satellite imagery provides multispectral data beyond visible RGB that reveal plant conditions.
- Combining satellite data with spatial embeddings can model the Earth as a sphere more accurately.
Satellite Data Boosts Species Models
- Satellite imagery and citizen science data can enhance traditional species distribution models.
- Satellite data offers finer resolution and global coverage than many traditional environmental variables.
Satellite Imagery Reveals Habitat
- Satellite imagery at 10-meter resolution offers more precise habitat info than traditional environmental data.
- It provides valuable habitat context even without resolving individual birds in images.