What happened
Google and NASA's JPL have developed an AI model called MAPL-EMIT, which uses data from NASA's EMIT satellite instrument to detect, quantify, and localize methane plumes globally. The model was trained on 3.6 million physics‑simulated methane plumes and detects 50% more plumes than human experts. It identified more than 23,000 additional plumes worldwide, including plumes at 24 of the world’s 25 largest‑emitting landfills.
The MAPL-EMIT model is an end‑to‑end vision transformer that processes the complete EMIT radiance spectrum to retrieve methane enhancements across all pixels in a scene jointly. It couples a Swin‑v2‑S transformer encoder with a convolutional decoder in a U‑Net‑like architecture and performs three tasks simultaneously: quantifying methane enhancement per pixel, delineating each plume’s shape and boundaries, and localizing each emission’s source.
The model was trained on 3.6 million synthetic plumes generated using a physics‑based simulation pipeline and injected into real EMIT scenes. Researchers partitioned 235,000 EMIT tiles into training, validation, and test sets and trained the model on 32 Google TPU chips for roughly 96 hours.
On real‑world benchmarks, MAPL‑EMIT captured 84 % of hand‑annotated NASA EMIT L2B plume complexes across a test set of 1,084 EMIT granules while identifying roughly 1.5 times as many plausible plumes as human analysts.
Why it matters
The AI model has significant implications for methane monitoring and reduction efforts. Methane is a potent greenhouse gas with a warming potential 30 times that of carbon dioxide over a 100‑year timeframe. The model can help identify and track methane emissions from various sources, including landfills, agriculture, and energy production. This information can be used to inform policy decisions and mitigation strategies to reduce methane emissions and combat climate change.
Methane’s high global warming potential makes accurate detection crucial for climate mitigation; the model improves plume detection rates by 50 % over human experts and uncovers thousands of previously unrecorded sources.
By pinpointing emissions from landfills, agriculture, and energy facilities, MAPL‑EMIT provides actionable data for regulators and companies aiming to meet the Global Methane Pledge’s 30 % reduction target by 2030.
The model’s finer spatial resolution (60 m) compared with existing satellite products enables source‑level attribution, supporting targeted repair and reduction efforts in high‑emitting sectors.
What to watch next
Future developments and applications of the MAPL-EMIT model, including its potential integration with other satellite instruments and its use in various industries. The model's ability to detect and quantify methane plumes could have significant implications for sectors such as oil and gas, agriculture, and waste management.
Integration of MAPL‑EMIT outputs into public Earth Engine tools and how policymakers and industry adopt the data for emission‑reduction programs.
Potential expansion of the approach to other greenhouse gases and multisatellite retrievals as NASA’s next‑generation imaging spectrometers come online.
Evaluation of false‑positive rates in complex terrain and subsequent refinements that could improve reliability for operational monitoring.