Independent validation of IASI/MetOp-A LMD and RAL CH ₄ products using CAMS model, in situ profiles, and ground-based FTIR measurements

Abstract. In this study, we carried out an independent validation of two methane retrieval algorithms using spectra from the Infrared Atmospheric Sounding Interferometer (IASI) that has been aboard the Meteorological Operational Satellite A (MetOp-A) since 2006. Both algorithms, one developed by the Laboratoire de Météorologie Dynamique (LMD), called the non-linear inference scheme (NLISv8.3), and the other by the Rutherford Appleton Laboratory (RAL), referred to as RALv2.0, provide long-term global CH4 concentrations using distinctively different retrieval approaches (neural network vs. optimal estimation, respectively). They also differ with respect to the vertical range covered, where LMD provides mid-tropospheric dry-air mole fractions (mtCH4), and RAL provides mixing ratio profiles from which we can derive total column-averaged dry-air mole fractions (XCH4) and potentially two partial column layers (qCH4). We compared both CH4 products using the Copernicus Atmospheric Monitoring Service (CAMS) model, in situ profiles (range extended using CAMS model data), and ground-based Fourier transform infrared (FTIR) remote-sensing measurements. The average difference (in mtCH4) with respect to in situ profiles for LMD ranges between −0.3 and 10.9 ppb, while for RAL the XCH4 difference ranges between −4.6 and −1.6 ppb. The standard deviation (SD) of the observed differences between in situ measurements and RAL retrievals is 14.1–21.9 ppb, which is consistently smaller than that between LMD retrievals and in situ measurements (15.2–30.6 ppb). By comparing with ground-based FTIR sites, the mean differences are within ±10 ppb for both RAL and LMD retrievals. However, the SD of the differences at the ground-based FTIR stations shows significantly lower values for RAL (11–15 ppb) than for LMD (about 25 ppb). The long-term trend and seasonal cycles of CH4 derived from the LMD and RAL products are further investigated and discussed. The seasonal variation in XCH4 derived from RAL is consistent with the seasonal variation observed by the ground-based FTIR measurements. However, the overall 2007–2015 XCH4 trend derived from RAL measurements is underestimated, if not adjusted, for an anomaly occurring on 16 May 2013 due to a L1 calibration change. For LMD, we see very good agreement at the (sub)tropics (<35° N–35° S) but notice deviations in the seasonal cycle (both in the amplitude and phase) and an underestimation of the long-term trend with respect to the RAL and reference data at higher-latitude sites.