Reducing Error in Airborne Imaging Spectrometer Retrieval of Greenhouse Gas Emissions Due to Geometric and Atmospheric Effects

Abstract:

Imaging spectroscopy has been used to retrieve enhanced methane concentrations and estimate fluxes from a variety of point source emitters. Retrieval techniques rely on matched filters to determine methane concentration-path length within image pixels. The “target” methane absorption spectrum required by matched filters is affected by a variety of geometric and atmospheric parameters; in other words, the change in absorption corresponding to a change in methane concentration is scene-dependent. Use of a static absorption spectrum will result in under- or over-estimation of methane enhancement and resulting fluxes. While an absorption spectrum for specific geometric and atmospheric conditions can be simulated using radiative transfer modeling, this approach is time-consuming and not practical for rapid processing of airborne campaign data.

To enable fast and portable generation of target absorption spectra for matched filter-based methane retrievals across a wide range of geometric and atmospheric conditions, and to improve retrieval accuracy, we use radiative transfer modeling to generate a table of absorption spectra at prespecified atmospheric conditions and utilize a linear spline interpolant to approximate the methane absorption spectra based on scene-specific parameters. Our method accounts for variation in solar zenith angle, ground altitude, sensor altitude, and column water vapor. Using 7 selected benchmark scenes from the AVIRIS-NG instrument that contain known point-source methane plumes, the improvement in retrieved methane concentration using this interpolation method versus generic-spectrum results is quantified. These scenes include 25 unique point-source emitters. The known methane sources include oil and gas facilities, agricultural, mining, and waste management facilities.

Use of scene-specific, interpolated absorption spectra in methane retrievals produced an average pixelwise error of 0.47±0.49% within plumes. In comparison, a standard HITRAN-based absorption spectrum produced an average pixelwise error of 15±5.1% within plumes. This new method reduces RMS error within plume pixels by an average of 273 ppm-m. These results demonstrate progress towards a more accurate method for retrieving methane emissions from airborne imaging spectroscopy.