Quantum computing for earth observation: monitoring methane emissions

Methane is a potent greenhouse gas with a 100-year global warming potential about 30 times greater than CO₂, and is responsible for roughly 30% of today’s global temperature rise. More than 60% of emissions are human-made, yet current monitoring is hindered by incomplete, fragmented, and uncertain data. Reconciling diverse data sources is therefore critical to achieving accurate and actionable methane inventories. Classical Earth observation approaches use low-resolution but frequent Sentinel-5P imagery to flag possible methane plumes, which are then analysed in higher-resolution Sentinel-2 images. Deep learning models, such as U-Nets, segment plumes and enable inversion for leak locations and emission rates. While effective, these methods remain computationally demanding and often struggle with the scale, heterogeneity, and uncertainty of global data.

This applied research project investigates whether quantum computing can offer advantages for methane monitoring. Two complementary angles are explored: a top-down approach to detect new leaks and a bottom-up approach to update emission inventories.

This report evaluated quantum-inspired methods for satellite-based methane-plume detection in a controlled setting. The results show that temporally combining multiple Sentinel-2 observations through quantum-inspired mechanisms can bring modest gains over a simple classical baseline, suggesting that new temporal fusion designs may add value. No computational quantum advantage was demonstrated, but the work identifies a promising direction that warrants further validation.