Retrieval of atmospheric CH₄profiles from Fourier transform infrared data using dimension reduction and MCMC

Abstract

We introduce an inversion method that uses dimension reduction for the retrieval of atmospheric methane (CH4) profiles. Uncertainty analysis is performed using the Markov chain Monte Carlo (MCMC) statistical estimation. These techniques are used to retrieve CH4 profiles from the ground‐based spectral measurements by the Fourier Transform Spectrometer (FTS) instrument at Sodankylä (67.4°N, 26.6°E), Northern Finland. The Sodankylä FTS is part of the Total Carbon Column Observing Network (TCCON), a global network that observes solar spectra in near‐infrared wavelengths. The high spectral resolution of the data provides approximately 3 degrees of freedom about the vertical structure of CH4 between around 0 and 40 km. We reduce the dimension of the inverse problem by using principal component analysis. Smooth and realistic profiles are sought by estimating three parameters for the profile shape. We use Bayesian framework with adaptive MCMC to better characterize the full posterior distribution of the solution and uncertainties related to the retrieval. The retrieved profiles are validated against in situ AirCore soundings which provide an accurate reference up to 20–30 km. The method is presented in a general form, so that it can easily be adapted for other applications, such as different trace gases or satellite‐borne measurements where more accurate profile information and better analysis of the uncertainties would be highly valuable.

Likelihood Informed Dimension Reduction for Remote Sensing of Atmospheric Constituent Profiles

“…To solve the inverse problem related to the FTIR measurement [16], we use adaptive MCMC [5,10] and SWIRLAB [15] toolboxes for Matlab. The results from newly implemented LIS-algorithm as well as from the previous prior reduction method are compared against a full dimensional MCMC simulation using the Hellinger distance of approximations to the full posterior.…”

“…However, there are about three degrees of freedom in the FTIR signal for the vertical profile information. To construct basis functions that could utilize this information a method that uses prior reduction was developed in [16]. It is based on the singular value decomposition on the prior covariance matrix,…”

“…This in effect makes the statistical estimation problem independent of the discretization of the inverse problem. In addition, we compare LIS to a dimension reduction method based on prior covariance matrix truncation used earlier [16].…”

A Singular Value Decomposition framework for retrievals with vertical distribution information from greenhouse gas column absorption spectroscopy measurements

“…Dimension reduction via SVD has been previously used both for satellite retrievals (Masiello et al, 2012;Thompson, 1992;5 Butz et al, 2010), ground-based spectrometers (Tukiainen et al, 2016) and laboratory laser absorption measurements (Bomse and Kane, 2006). The SVD approach described here comes closest to the one applied for satellite methane retrievals (Butz et al, 2010), but performs the retrieval in the principal component basis to eliminate bias originating from the choice of the uninformative prior used (see section 3.5).…”

A scientific algorithm to simultaneously retrieve carbon monoxide and methane from TROPOMI onboard Sentinel-5 Precursor

“…Outliers with higher values compared to TCCON are more rare and dominated by a handful of 30 collocations at East Trout Lake. This exceptional lack of XCH 4 agreement occurs on four days in the time period February 10-21 as well as on March 29 and may be attributable to Arctic polar vortex air above East Trout Lake potentially causing the following related issues: associated fronts of different air masses may complicate the identification of collocations near the vortex edge and/or the stratospheric part of the methane profile may be largely affected by the polar vortex leading to a considerable deviation from the assumed apriori profile shapes (Tukiainen et al, 2016). It is verified that the impact of outliers on the regression is marginal by repeating the fit with the Huber linear regression model (Huber and Ronchetti, 2009), which is robust to outliers and provides similar results to the standard linear regression here.…”