Google and NASA Jet Propulsion Laboratory (JPL) have released MAPL-EMIT, a Vision Transformer model designed to detect methane plumes from space, disentangle overlapping emissions and trace each plume back to its source.
How it improves on earlier methods
NASA's EMIT instrument has been used to monitor methane from orbit, but its matched-filter detection approach reportedly missed roughly half of the plumes and could not reliably separate overlapping sources. MAPL-EMIT was trained on 3.6 million synthetic plume scenarios to address these limitations.
In evaluations presented by the developers, MAPL-EMIT achieved 84% recall and detected about 50% more plumes than NASA's matched-filter benchmark. The model can separate overlapping plumes and assign them to individual facilities, allowing clearer attribution of emissions.
Performance and openness
In testing, the system identified 24 of the world's 25 largest landfills. Google and JPL have made the model, the training data and the code openly available, enabling independent verification and further research.
Why this matters
Methane is a much more powerful short-term greenhouse gas than carbon dioxide: the reported figures cite roughly 30 times the potency of CO2, and attribute about 25% of current global warming to methane. Reducing methane emissions is therefore a high-impact near-term climate strategy.
Previously, operators could often dispute that a detected plume originated from their facility because existing satellite methods could not reliably separate nearby sources. MAPL-EMIT’s ability to disentangle overlapping plumes and attribute them to specific sites changes that dynamic and allows detections to be publicly documented.
This capability is particularly relevant for monitoring commitments such as the Global Methane Pledge, signed by 125 countries, because continuous, automated and openly available satellite-based detection increases the potential for accountability.
Practical implications
By sharpening satellite-based methane detection, MAPL-EMIT makes many previously hard-to-see leaks observable and attributable. The open release of model, data and code means regulators, researchers and civil society can inspect detections and potentially act more quickly on identified emission sources.
Summary
MAPL-EMIT represents a significant advance in methane monitoring from space: it finds more plumes, separates overlapping emissions and enables source attribution, while being released openly to support verification and accountability in global methane reduction efforts.
