The Findings From WEF & MIT’s AI Earth Observation Research

Climate-driven disasters are increasing in frequency and severity, yet the technologies designed to monitor and predict them are struggling with slow processing times and limited accessibility. 

The World Economic Forum (WEF), the international organisation based in Geneva, has now published research in partnership with the MIT Media Lab, examining how AI is transforming earth observation from a data collection exercise into a real-time climate intelligence system.

The whitepaper, titled Charting the Future of Earth Observation: Technology Innovation for Climate Intelligence, makes the case that AI and machine learning (ML) are finally unlocking the potential of satellite data that has, until now, sat largely unused due to prohibitive processing times. 

Earth observation refers to the collection and analysis of data on Earth’s physical, chemical and biological systems, primarily through satellite-based remote sensing.

“This white paper, written in collaboration with the Massachusetts Institute of Technology (MIT) Media Lab, highlights the transformative potential of EO for climate intelligence and forecasting,” say Sebastian Buckup, Managing Director at the WEF and Dava Newman, Director at MIT’s Media Lab. 

“By combining the research capabilities of the MIT Media Lab with the global platform of the World Economic Forum, the paper identifies technology pipelines accelerating the processing and analysis of satellite EO data to provide unparalleled insights into climate change,” they say.

How Microsoft Aurora and NASA Prithvi accelerate predictions

By 2032, satellite earth observation systems are forecast to generate more than two exabytes of data, equivalent to more than two billion gigabytes, accounting for 86% of all data from the space applications segment. 

The research indicates that more than half of the variables needed to measure climate conditions can only be accurately captured from space, making AI-powered processing not just useful but critical for climate action.

The whitepaper positions ML as the breakthrough that changes everything. 

These AI systems, which improve through exposure to data, can now process earth observation datasets in near real-time, delivering predictions up to 1,000 times faster than previous techniques. 

That shift turns satellite images into actionable insights in minutes rather than weeks.

Meanwhile foundation models, which are large AI systems trained on diverse datasets, enable localised and global forecasting at speeds that traditional physics-based models simply cannot match. 

As a result, Microsoft Corporation’s Aurora system can predict air pollutant levels worldwide in seconds, while NASA’s open-source Prithvi-weather-climate model supports both flood mapping and crop yield projections. 

The practical applications are already visible. 

Following Hurricane Beryl in 2024, Microsoft partnered with Planet Labs, a satellite imaging company based in San Francisco, to assess and map building damage in Carriacou, Grenada. 

ML models analysed satellite imagery at the pixel level, categorising damage extent and severity across buildings in hours, work that would traditionally have taken weeks or months.

New low Earth orbit satellite constellations, such as the planned Muon Space system, will deliver near real-time data capable of detecting fire ignition sites as small as 25 square metres, with a revisit time of just 20 minutes. 

Low Earth orbit refers to satellites positioned between 160 and 2,000 kilometres above Earth’s surface. 

When combined with AI processing that can identify anomalies in orbit before data even reaches the ground, the speed advantage becomes substantial.

The hardware side is advancing too. 

The Landsat Next mission, scheduled to launch in 2030, will collect 26 superspectral bands, more than double the number captured by previous generations. 

But without AI to process that volume of data, the additional detail would be largely academic.

The impact of MIT’s Media Lab and Digital Earth Africa democratising climate data

AI is also making earth observation accessible beyond technical experts and well-resourced institutions. 

The MIT Media Lab’s Earth Mission Control and Africa’s Digital Earth Africa initiative use open-source technology to provide water monitoring and flood forecasting to hundreds of thousands of sites across Africa. 

Digital Earth Africa is a programme managed by Geoscience Australia, the government agency responsible for geoscience research.

The European Space Agency, the intergovernmental organisation coordinating space activities for European states, plans to develop a digital earth observation assistant using Gen AI – systems that can create new content and responses based on training data. 

The tool will answer text-based queries and provide visual insights from earth observation data, translating complex satellite feeds into information that policymakers and communities can actually use.

However, the challenges remain substantial. 

Data interoperability requires standardisation across platforms and formats. Infrastructure access demands investment in cloud-based processing and digital literacy programmes for vulnerable communities. 

And as the technology evolves, workforce training needs expand across AI, satellite engineering and geospatial analytics.

“The paper also highlights the need for accessible and inclusive climate insights, especially for communities most vulnerable to the effects of climate change,” Sebastian and Dava say.