Satlas
Satlas is an AI tool developed by AI2 (Allen Institute for AI) that allows users to explore and analyze changes happening on Earth through ...
Last verified:
What is Satlas?
Satlas is a platform developed by the Allen Institute for AI (AI2) for visualizing and downloading global geospatial data products generated by AI using satellite images. The platform combines modern AI with publicly available satellite imagery from the European Space Agency's Sentinel-2 satellites to provide monthly monitoring of the planet at a global scale.
Currently, Satlas includes three main data products that are updated monthly: Marine Infrastructure (positions of offshore wind turbines and platforms), Renewable Energy Infrastructure (positions of onshore wind turbines and solar farms), and Tree Cover (tree canopy coverage percentage of each 100 m² cell). The data is available as monthly maps starting from January 2016, with marine and renewable energy infrastructure published as GeoJSON files and tree cover as GeoTIFFs.
Satlas is designed for researchers, environmental monitors, policymakers, urban planners, and anyone interested in planetary and environmental monitoring. The platform features a super-resolution explorer that visualizes AI-generated high-resolution imagery (4x enhancement from original Sentinel-2 resolution) on a global scale. All data, training data, and model weights are freely available for download for offline analysis under an open license (ODC-BY).
The platform uses high-accuracy deep learning models trained on manually labeled data including 36,000 wind turbines, 4,000 solar farms, 7,000 offshore platforms, and 3,000 tree cover canopy percentages. These models leverage foundation models pre-trained on SatlasPretrain, a large-scale remote sensing dataset with 302 million labels across over a hundred tasks.
Satlas pricing
Pricing model: Free
Satlas is completely free to use. All geospatial data products, training data, and model weights are available for download at no cost. The data is released under an open license (ODC-BY), which allows free use, distribution, and modification with attribution. There are no paid plans, subscription tiers, or premium features. Users can visualize data in the Satlas Map or download GeoJSON files (for marine and renewable energy infrastructure) and GeoTIFF files (for tree cover) for offline analysis.
Satlas pros
- Fully free and open access to all geospatial data
- Monthly updates providing timely data
- Global coverage of all land masses
- Data available since January 2016 for historical analysis
- Open license (ODC-BY) allows unrestricted use
- Downloadable GeoJSON and GeoTIFF formats for offline analysis
- Super-resolution explorer with 4x AI-enhanced image resolution
- Human-level accuracy in detecting wind turbines and solar farms
- Training data and model weights freely available
- Foundation models pre-trained on 302 million labels
- Three distinct data products for energy and environmental monitoring
- Sentinel-2 imagery provides weekly global coverage
- No payment or subscription required
- Suitable for renewable energy tracking and emissions reduction
- Can be used for disaster relief and urban planning applications
Satlas cons
- Only three data products currently available
- 10 m/pixel resolution from Sentinel-2 is relatively low
- Tree cover data limited to 100 m² cell percentages
- No mobile app available
- Limited to offshore/onshore wind and solar, no other energy types
- No real-time data, only monthly updates
- Super-resolution accuracy still being improved and quantified
- No API documented for programmatic access
- Data products planned but not yet implemented (crop types, urban land use)
- Interface may require technical knowledge for offline analysis
Frequently asked questions about Satlas
What is Satlas?
Satlas is a platform for exploring global geospatial data generated by AI from satellite imagery. It provides monthly updates on marine infrastructure (offshore wind turbines and platforms), renewable energy infrastructure (onshore wind turbines and solar farms), and tree cover coverage, all visualized on an interactive map and available for download.
What satellite imagery does Satlas use?
Satlas uses satellite images from the European Space Agency's Sentinel-2 satellites. These images are publicly available, have a resolution of 10 m/pixel, and capture the bulk of Earth's land mass weekly. Satlas processes these images monthly to update its geospatial data products.
How often is Satlas data updated?
Satlas data is updated on a monthly basis. Every month, the platform downloads new satellite images covering Earth's entire land mass and applies deep learning models to derive an up-to-date global snapshot of each geospatial data product. Data is available as monthly maps starting from January 2016.
What data formats are available for download?
Marine infrastructure and renewable energy infrastructure data are published as GeoJSON files for each month. Tree cover data is published as a set of GeoTIFFs for each month. All data can be downloaded for offline analysis from the Satlas website.
What is the Super-Resolution Explorer?
The Super-Resolution Explorer is a feature on Satlas that uses deep learning to generate high-resolution images from multiple low-resolution Sentinel-2 images of the same location captured at different times. It enhances the original satellite imagery by 4x resolution, allowing users to explore the world with AI-generated detailed visuals.
Is Satlas data free to use?
Yes, all Satlas data is completely free. The data is released under an open license (ODC-BY), which allows free use, distribution, and modification with proper attribution. This includes the geospatial data products, training data (36K wind turbines, 4K solar farms, 7K offshore platforms, 3K tree cover labels), and model weights.
What is SatlasPretrain?
SatlasPretrain is a large-scale remote sensing dataset containing several terabytes of Sentinel-2 images with 302 million labels across over a hundred tasks. Foundation models are pre-trained on this dataset to perform tasks like land cover segmentation, crop type classification, and building detection. These foundation models provide better and more consistent performance than models trained from scratch.
Who should use Satlas?
Satlas is useful for researchers, environmental monitors, policymakers, urban planners, and organizations working on emissions reduction, disaster relief, renewable energy tracking, and planetary monitoring. It is particularly valuable for tracking renewable energy infrastructure growth across political boundaries and for applications requiring timely global geospatial data.
What are the accuracy levels of Satlas models?
The deep learning models in Satlas have been trained to extract data like wind turbine positions from satellite imagery with human-level accuracy. The models were trained on manually labeled examples including 36,000 wind turbines, 4,000 solar farms, and 7,000 offshore platforms, achieving precision comparable to human analysis.
What data products are planned for the future?
Satlas plans to add more geospatial data products over time. Currently exploring models for mapping urban land use, crop types, and land cover, with hopes to incorporate a subset by the end of 2023. In the long term, they plan to release tools making it easier for other teams to build similar geospatial data products, including annotating examples, training models, and deploying them.