Abstract Description: Canada’s upstream oil and gas sector accounts for approximately 44% of anthropogenic methane emissions, which must be reduced by 75% of 2012 levels by 2030. To achieve this goal, industry and regulators require instrumentation that can detect and quantify methane emissions. In this regard, airborne hyperspectral imaging over the long-wavelength infrared (LWIR) spectrum holds significant promise. This study assesses the Telops Hyper-Cam Airborne Mini xLW through controlled release measurements. This instrument generates a data cube comprised of a programmable number of pixels (up to 320 x 256), where each pixel contains an absolute radiance spectrum between 7.4-11.8 um. Under normal operating conditions, measurements are obtained at a spectral resolution of 10 cm-1 and an altitude of 350 m, corresponding to a ~30 m wide inspection window.
Data cubes are analyzed using K-means clustering and clutter-matched filtering to identify and binarize the plume, followed by a spectroscopic model to calculate the methane column density for each pixel. Finally, column densities are converted into a pixel mass-map, which is combined with a plume advection model and wind speed to obtain an emissions rate estimate.
The performance of the system is assessed through controlled release measurements made over bare and snow-covered ground. Initial results showed that the Hyper-Cam inferred releases systematically underpredicted the known release rates, which was attributed to images that did not adequately capture the entire plume. Once these images were identified using a convolutional neural network and removed from the dataset, the bias was largely eliminated.
