Detection of methane plumes in hyperspectral images from SENTINEL-5P by coupling anomaly detection and pattern recognition

Reducing methane emissions is essential to tackle climate change. Here, we address the problem of detecting large methane leaks using hyperspectral data from the Sentinel-5P satellite. For that we exploit the fine spectral sampling of Sentinel-5P data to detect methane absorption features visible in the shortwave infrared wavelength range (SWIR). Our method involves three separate steps: i) background subtraction, ii) detection of local maxima in the negative logarithmic spectrum of each pixel and iii) anomaly detection in the background-free image. In the first step, we remove the impact of the albedo using albedo maps and the impact of the atmosphere by using a principal component analysis (PCA) over a time series of past observations. In the second step, we count for each pixel the number of local maxima that correspond to a subset of local maxima in the methane absorption spectrum. This counting method allows us to set up a statistical a contrario test that controls the false alarm rate of our detections. In the last step we use an anomaly detector to isolate potential methane plumes and we intersect those potential plumes with what have been detected in the second step. This process strongly reduces the number of false alarms. We validate our method by comparing the detected plumes against a dataset of plumes manually annotated on the Sentinel-5P L2 methane product.