Satellites
Satellites are a powerful tool for simultaneously monitoring animals, and their surrounding environments. A single very high resolution (VHR) satellite image can cover hundreds of square kilometers, with a pixel representing less than 50cm of the ocean’s surface. At this resolution, large animals such as whales turtles and seals can be seen in an image.
Typically, we identify an area to survey and work with commercial satellite providers to collect imagery over this area. Government agencies, such as NASA and ESA, may also collect data, or provide archived satellite imagery over a site of interest. With this imagery, we can manually comb through a scene to identify and count animals.
The massive scale of satellite imagery makes manual annotation inefficient, so we use the animals that we find as training data for automated AI detectors that will allow animals to be counted at scale. By rapidly and efficiently counting animals in remote regions across the globe, we can receive real-time information on how animal populations are faring in terms of abundance. These new tools will allow us to monitor key populations in areas that are difficult to access and survey.
How it Works:
-
Satellite imagery is acquired and processed by an image provider.
-
Human annotators systematically examine the satellite imagery and identify megafauna.
-
Image chips of animals are extracted from the broader scene to train AI models.
-
Sightings are verified and assessed for quality, producing animal counts across time and region
-
Animal count data is paired with behavioral tag data and environmental information to model population abundance.
