Objectives of the service
Project objectives
The project develops and tests a satellite-based approach for identifying significant methane emission events around critical energy infrastructure. The approach combines Earth Observation imagery with established spectral methods and an artificial intelligence / machine learning (AI/ML) segmentation workflow. The Proof of Concept assesses how reliably these methods operate under realistic European conditions, using stakeholder-relevant scenarios in the Czech Republic and global reference observations. The work also clarifies where satellite monitoring can complement existing inspection practices and where current sensor sensitivity, cloud conditions and satellite overpass timing limit operational value.
Users and their needs
Primary users are operators and asset managers responsible for methane-emitting gas infrastructure. The Feasibility Study used NET4GAS in the Czech Republic as the reference stakeholder. Key needs identified during stakeholder engagement include:
-
independent information that complements existing methane monitoring
-
better visibility of selected high-volume operational events
-
support for operational review and potential compliance-related reporting
-
less reliance on frequent, costly and logistically demanding aerial campaigns
Target country: Czech Republic.
Service/ system concept
Service/system concept
The concept processes satellite imagery over defined areas and times, combines a physics-based methane baseline with AI/ML plume segmentation, and produces maps, detection indicators and supporting diagnostics. The workflow starts with satellite observations and user-provided location/event information, then filters imagery for suitable conditions, analyses methane-sensitive spectral changes and compares results against reference observations. The PoC implemented the workflow as a reproducible, script-based pipeline. For a future operational service, outputs could be presented through a secure digital environment with visualisation, alerts and analytics, subject to validation for a technically suitable use case.
Space Added Value
Earth Observation adds wide-area, repeatable remote coverage without dedicated local survey flights. Sentinel-2 provides a nominal 3–5 day revisit, while Landsat 8/9 and commercial WorldView-3 add complementary spatial and acquisition characteristics. The PoC showed that these assets can support methane plume analysis for sufficiently strong emissions, but practical monitoring frequency is lower than nominal revisit because cloud cover, scene quality and event timing matter. The project therefore positions satellite monitoring as a complementary information layer rather than a replacement for all field or aerial methods.
Current Status
Current status
The project reached Final Review in September 2026. The Proof of Concept implemented both the AI/ML and traditional methane-detection branches, created approximately 41,000 training samples, and extended the real-reference archive to 4,424 Sentinel-2 temporal pairs. After suitability screening, approximately 2,320 pairs were retained and 2,318 were successfully evaluated. The best synthetic-data-trained model achieved a validation selection score of 0.6344 at a probability threshold of 0.95. On real reference observations, 1,306 predictions showed non-zero spatial overlap with reference plume geometries (56.3%); mean IoU was 0.204 and median IoU 0.090. NET4GAS events produced no confirmed methane plume detections, and a reported persistent emission of about 4 kgCH4/h was not detected in WorldView-3 imagery.
