Geospatial data analysis promises to revolutionize the way agriculture, urban planning and disaster relief is done — and thanks to a variety of projects that make use of artificial intelligence, Microsoft and Seattle’s Allen Institute for AI are part of that revolution.
The Allen Institute for AI, also known as AI2, recently rolled out Satlas, a new software platform for exploring global geospatial data generated from satellite imagery. Meanwhile, Microsoft’s AI for Good Lab is working with public and private institutions in Colombia on Project Guacamaya, which uses AI tools to monitor and understand conditions in the Amazon Rainforest.
Satlas: Snapshots of global development
Satlas starts with images from the European Space Agency’s twin Sentinel-2 satellites. A deep-learning computer model analyzes the images to identify features such as offshore platforms, solar farms and the amount of tree cover in a given area. Those features are then displayed on a zoomable online map that can track changes over time.
Training the model was no easy task. “We have manually labeled 36K wind turbines, 4K solar farms, 7K offshore platforms and 3K tree cover canopy percentages in Sentinel-2 imagery,” AI2 research scientist Fayven Bastani said in a blog post explaining the project.
The first three data products focus on the positions of offshore wind turbines and platforms, the locations of solar farms and onshore wind turbines, and the percentage of tree canopy coverage for 100-square-meter plots of land. “Over time, we plan to release additional geospatial data products,” Bastani said.
Satlas’ developers have already posted some sample applications — for example, showing how commercial logging has reduced tree cover on Washington’s Olympic Peninsula, or how heavily Brazil has invested in solar farms over the past few years.
The Satlas team has also developed AI tools for sharpening the resolution of satellite imagery by a factor of four, from 10 meters to 2.5 meters per pixel.
“We use deep learning models to generate a high-resolution image from many low-resolution images of the same location captured at different times,” Bastani explained. “The model tries to combine information across low-resolution images to predict sub-pixel details.”
Another AI2 project, Skylight, is already looking into how Satlas’ marine infrastructure data could be used to improve its system for tracking vessels at sea, including ships that may be involved in illegal fishing operations. “We are actively looking for other use cases,” Bastani said.
In the months ahead, the Satlas team hopes to expand its line of geospatial data products to include mapping tools for urban land use, agricultural crop types and land cover.
Project Guacamaya: Protecting the Amazon
With Project Guacamaya, Microsoft and its partners are addressing a South American issue with global implications: the deforestation of the Amazon Rainforest.
The rainforest is known as the “lungs of the planet,” thanks to its role in global climate regulation. But those lungs are facing increasing threats from illegal logging and mining. Experts documented a 21% increase in the rate of Amazon deforestation in 2022.
“We need to be using technology and innovation to think outside of the box to face the problems that we have,” Diego Ochoa, director of external affairs at the Alexander von Humboldt Institute in Colombia, said in a Microsoft blog posting about the project. “We have powerful tools at hand to promote change in society.”
Project Guacamaya (which takes its name from the Mayan word for “macaw”) uses an array of AI models for satellite image analysis, plus AI models that have been adapted from existing Microsoft projects for camera trap analysis and bioacoustics. The project blends data from Planet Labs’ satellites, hidden cameras and acoustic sensors to track the health of the rainforest and the species that live there.
Colombian conservationists can use the data from Project Guacamaya to sound the alarm about specific environmental threats.
“This is not just about monitoring, and it’s not just about artificial intelligence or even generating a scientific paper,” Ochoa said. “We need to inspire society, and this kind of project, using technology, using innovation, using AI, hopefully will inspire Colombians to change the way they think about biodiversity and nature. This will help people get closer to what we have in those areas.”
In addition to the Humboldt Institute and Planet Labs, Microsoft’s collaborators in Project Guacamaya include the Sinchi Amazonian Scientific Research Institute and CINFONIA, an AI research center at the University of the Andes.
In Brazil, Microsoft is working with Imazon and the Fundo Vale Foundation to develop AI models for another Amazon-monitoring project called PrevisIA. And that’s just the start.
“All of the models we’re going to be doing will be open source,” said Juan Lavista Ferres, vice president and chief data scientist at the AI for Good Lab. “Ideally, once we finish these models, we can look at Ecuador, we can look at Peru and say, ‘We have this. Can we sync and see how you can use it?’”
For more perspectives on geospatial data analysis and global environmental monitoring, check out what Amazon and BlackSky are up to.
