Artificial intelligence is emerging as a transformative force in understanding and protecting our planet’s ecosystems. As human demands for land and resources intensify, AI offers powerful tools to help governments, companies, and conservation groups collect field data, generate new insights, and translate those insights into meaningful action. Today, we’re excited to share groundbreaking developments in three key areas where AI is making a significant impact on environmental conservation.
Predicting Deforestation with Satellite Precision
Forests represent one of our biosphere’s most critical components—storing carbon, regulating rainfall, mitigating floods, and hosting the majority of Earth’s terrestrial biodiversity. Despite their importance, forests continue to disappear at alarming rates. While satellite-based deforestation tracking has existed for over two decades, recent advancements are taking monitoring to unprecedented levels.
In collaboration with the World Resources Institute, researchers have developed a sophisticated model that analyzes the drivers of forest loss—from agriculture and logging to mining and fire—at an incredible 1km² resolution, covering data from 2000-2024. The newly released benchmark dataset for predicting deforestation risk uses pure satellite inputs, eliminating the need for specific local data layers like roads. Built around vision transformers, this efficient model architecture delivers accurate, high-resolution predictions down to 30 meters across large regions.
Mapping Earth’s Species Distribution
To protect threatened species, we must first know where they live. With over 2 million known species and millions more awaiting discovery, this represents a monumental challenge. Google researchers are addressing this through an innovative AI-powered approach that produces species range maps at unprecedented scale—covering more species across more of the world at higher resolution than ever before.
The Graph Neural Net (GNN) model combines open databases of field observations with satellite embeddings from AlphaEarth Foundations and species trait information such as body mass. This integrated approach allows researchers to infer likely geographical distributions for multiple species simultaneously, while enabling scientists to refine these distributions with additional local data and expertise.
In a pilot project with QCIF and EcoCommons, the model has been used to map Australian mammals like the Greater Glider—a nocturnal, fluffy-tailed marsupial inhabiting old-growth eucalyptus forests. Twenty-three of these species maps are now available through the UN Biodiversity Lab, providing valuable resources for conservation planning.
Revolutionizing Field Monitoring Through Bioacoustics
All ecosystem understanding ultimately depends on field monitoring, and AI is revolutionizing this traditionally difficult and costly process. Bioacoustics presents a particularly compelling application, as birds, amphibians, insects, and other species use sound for communication, making audio an excellent modality for identifying species and assessing ecosystem health.
While reliable bioacoustic monitors are readily available, they generate vast audio datasets filled with unknown and overlapping sounds that are too large for manual review yet challenging for automated analysis. To address this complexity, researchers have released Perch 2.0—an updated animal vocalization classifier that represents the state of the art in bird identification.
This foundational model allows field ecologists to quickly adapt the system to identify new species and habitats anywhere on Earth. The technology is already making a difference through work with the University of Hawai`i, where Perch is guiding protective measures for endangered honeycreepers and analyzing juvenile calls to understand population health.
Integrating Models for Comprehensive Understanding
The ultimate goal of these initiatives is to empower decision-makers at all levels to take effective action in protecting our planet. Better data only leads to better decisions when it thoroughly captures what’s happening across entire ecosystems. That’s why researchers are working to integrate multiple models, combining data from satellites, images, bioacoustics, documents, and other modalities.
This comprehensive approach also incorporates models of human activity—including land-use changes, agricultural practices, yields, flood prevention, and other human-relevant consequences. By providing policymakers with a complete understanding of biosphere threats, we can help them take action to protect future generations of plants, animals, and people. If we can accurately model our environment, we can help it thrive.
These remarkable advancements represent just the beginning of AI’s potential in environmental conservation. As technology continues to evolve, we can expect even more sophisticated tools to emerge, offering new hope for preserving Earth’s precious biodiversity and ecosystems for generations to come.