Species distributions modeled worldwide using satellite data and huge observations
OCELOT: Planetary-Scale Species Distribution Modeling with Satellite Foundation Embeddings
Machine Learning
Summary
Species sightings are often recorded mostly near places where people live, making it hard to guess where animals live in remote areas. The authors created a new AI model called OCELOT that uses satellite images combined with billions of records to predict where species live, even in places rarely visited by people. Their approach better handles data biases and shares information across related species to improve predictions, especially for rare animals. Tests show it predicts species locations more accurately in distant regions than previous methods while keeping strong accuracy closer to known areas.
What this means in practice
- •For wildlife monitoring teams: Use improved species distribution maps to help identify animals captured on remote camera traps or audio recorders.
- •For conservation planners: Apply better geographic extrapolation of species habitats to plan protected areas in regions lacking direct observations.
Authors
Bart van Merriënboer, Lauren Harrell, Keith Anderson, Burcu Karagol Ayan, Elise Kleeman, Tom Denton, Drew Purves
Abstract
Large-scale repositories of species observations now contain billions of occurrence records from around the world, but these sightings cluster heavily in human-accessible areas. Existing models use only a small fraction of these records and are not evaluated on extrapolating to remote, unvisited regions--where species distribution models can be used as geographic priors to help classify data from camera traps and acoustic recorders. We scale relative occurrence species distribution models by training a multimodal implicit neural representation network called OCELOT that fuses spatial coordinates with satellite foundation embeddings on two orders of magnitude more observations than previous work: 2 billion GBIF records across 100,000 taxa. Scaling to this regime requires co-designing the data pipeline, taxonomic representation, loss formulation, and model architecture: spatiotemporal thinning and sub-linear cell sampling correct for spatial redundancy and observer bias, while an additive taxonomic readout shares representations across species, genera, and families to improve predictions for rare taxa. To evaluate performance in remote areas, we benchmark geographic extrapolation using the NDCG metric on spatial block holdouts up to 600 km. Satellite foundation embeddings improve extrapolation by 51% over coordinate-only models and outperform classical bioclimatic rasters. The resulting model also matches or outperforms existing baselines on established interpolation benchmarks, demonstrating that optimizing for geographic extrapolation does not compromise in-distribution accuracy.