Comparing imaging processing techniques with physical based inverse radiative transfer models for methane and carbon dioxide point emissions

Satellite remote sensing techniques offer the possibility of an independent global monitoring of carbon dioxide (CO₂) and methane (CH₄) emissions. Recently, point sources such as oil and gas facilities and power plants, which emit a high concentration of greenhouse gases (GHG) locally, have received particular attention. Therefore, present and upcoming satellite missions focus on collecting GHG concentration images with a high spatial resolution. The PRISMA (PRecursore IperSpettrale della Missione Applicativais) spaceborne imaging spectrometer is an Italian satellite which has been launched on March 22, 2019. It is the first satellite which provides open access hyperspectral images of backscattered sunlight with a spatial resolution of 30x30 meter and a spectral resolution of around 11nm. The measured absorption spectra in the shortwave infrared range cover strong CO₂ and CH₄absorption bands. Various methods can be used to retrieve 2-dimensional CO₂ and CH₄fields above localized GHG sources.

Here, we compare data-driven and physics-based retrieval methods in application to PRISMA measurements above localized GHG sources such as oil and gas production facilities in Turkmenistan for CH₄ and coal-fired power plants for CO₂. The data-driven methods are variants of the matched filter technique while the physics-based methods built on spectroscopic radiative transfer modeling. While matched filter techniques use the spatial covariance of the observed scene, traditional physics-based retrievals operate on individual spectra without considering the two-dimensional scene. For a few cases, we examine the differences between both methods and conclude on strengths and weaknesses.