Dr Fei Yao

Active research projects:

Harnessing satellite observations of methane to inform climate mitigation strategies

Methane is a potent greenhouse gas that has large human-driven emissions and breaks down in the atmospheric much quicker than carbon dioxide. Reducing methane emissions therefore represents an important opportunity for short-term climate change mitigation. Satellite observations can now pinpoint hotspots of atmospheric methane on a spatial scale of tens of meters that reflect large point sources. We propose to develop trusted methane data products, underpinned by domain-level expertise, by applying novel machine learning (ML) algorithms to the growing data volumes of satellite observations of atmosphere methane.

We are developing a framework that can automatically detect near-real-time hotspots of anomalous atmospheric methane across the globe using Sentinel-2 satellite data. First, we prepare a labelled dataset of Sentinel-2 imagery with and without anomalous atmospheric methane. We use this dataset to train and evaluate a ML model that can identify anomalous atmospheric methane in Sentinel-2 images. We then apply the resulting ML model to detect hotspots of atmospheric methane from all Sentinel-2 observations. Finally, we translate our detections of atmospheric methane hotspots into emission estimates and uncertainties that can inform mitigation strategies.