Search

Recent advances in satellite observations of methane provide increased opportunities for inverse modeling. However, challenges exist in the satellite observation optimization and retrievals for high latitudes. In this study, we examine possibilities and challenges in the use of the total column averaged dry-air mole fractions of methane (XCH4) data over land from the TROPOspheric Monitoring Instrument (TROPOMI) on board the Sentinel 5 Precursor satellite in the estimation of CH4 fluxes using the CarbonTracker Europe-CH4 (CTE-CH4) atmospheric inverse model. We carry out simulations assimilating two retrieval products: Netherlands Institute for Space Research’s (SRON) operational and University of Bremen’s Weighting Function Modified Differential Optical Absorption Spectroscopy (WFM-DOAS). For comparison, we also carry out a simulation assimilating the ground-based surface data. Our results show smaller regional emissions in the TROPOMI inversions compared to the prior and surface inversion, although they are roughly within the range of the previous studies. The wetland emissions in summer and anthropogenic emissions in spring are lesser. The inversion results based on the two satellite datasets show many similarities in terms of spatial distribution and time series but also clear differences, especially in Canada, where CH4 emission maximum is later, when the SRON’s operational data are assimilated. The TROPOMI inversions show higher CH4 emissions from oil and gas production and coal mining from Russia and Kazakhstan. The location of hotspots in the TROPOMI inversions did not change compared to the prior, but all inversions indicated spatially more homogeneous high wetland emissions in northern Fennoscandia. In addition, we find that the regional monthly wetland emissions in the TROPOMI inversions do not correlate with the anthropogenic emissions as strongly as those in the surface inversion. The uncertainty estimates in the TROPOMI inversions are more homogeneous in space, and the regional uncertainties are comparable to the surface inversion. This indicates the potential of the TROPOMI data to better separately estimate wetland and anthropogenic emissions, as well as constrain spatial distributions. This study emphasizes the importance of quantifying and taking into account the model and retrieval uncertainties in regional levels in order to improve and derive more robust emission estimates.

Nitrous oxide (N2O) is the third most abundant anthropogenous greenhouse gas (after carbon dioxide and methane), with a long atmospheric lifetime and a continuously increasing concentration due to human activities, making it an important gas to monitor. In this work, we present a new method to retrieve N2O concentration profiles (with up to two degrees of freedom) from each cloud-free satellite observation by the Infrared Atmospheric Sounding Interferometer (IASI), using spectral micro-windows in the N2O ν3 band, the Radiative Transfer for TOVS (RTTOV) tools and the Tikhonov regularization scheme. A time series of ten years (2011–2020) of IASI N2O profiles and integrated partial columns has been produced and validated with collocated ground-based Network for the Detection of Atmospheric Composition Change (NDACC) and Total Carbon Column Observing Network (TCCON) data. The importance of consistency in the ancillary data used for the retrieval for generating consistent time series has been demonstrated. The Nitrous Oxide Profiling from Infrared Radiances (NOPIR) N2O partial columns are of very good quality, with a positive bias of 1.8 to 4% with respect to the ground-based data, which is less than the sum of uncertainties of the compared values. At high latitudes, the comparisons are a bit worse, due to either a known bias in the ground-based data, or to a higher uncertainty in both ground-based and satellite retrievals.

This paper provides an overview of the validation of National Oceanic and Atmospheric Administration (NOAA) operational retrievals of atmospheric carbon trace gas profiles, specifically carbon monoxide (CO), methane (CH4) and carbon dioxide (CO2), from the NOAA-Unique Combined Atmospheric Processing System (NUCAPS), a NOAA enterprise algorithm that retrieves atmospheric profile environmental data records (EDRs) under global non-precipitating (clear to partly cloudy) conditions. Vertical information about atmospheric trace gases is obtained from the Cross-track Infrared Sounder (CrIS), an infrared Fourier transform spectrometer that measures high resolution Earth radiance spectra from NOAA operational low earth orbit (LEO) satellites, including the Suomi National Polar-orbiting Partnership (SNPP) and follow-on Joint Polar Satellite System (JPSS) series beginning with NOAA-20. The NUCAPS CO, CH4, and CO2 profile EDRs are rigorously validated in this paper using well-established independent truth datasets, namely total column data from ground-based Total Carbon Column Observing Network (TCCON) sites, and in situ vertical profile data obtained from aircraft and balloon platforms via the NASA Atmospheric Tomography (ATom) mission and NOAA AirCore sampler, respectively. Statistical analyses using these datasets demonstrate that the NUCAPS carbon gas profile EDRs generally meet JPSS Level 1 global performance requirements, with the absolute accuracy and precision of CO 5% and 15%, respectively, in layers where CrIS has vertical sensitivity; CH4 and CO2 product accuracies are both found to be within ±1%, with precisions of ≈1.5% and ⪅0.5%, respectively, throughout the tropospheric column.

The impact of SEOM–IAS (Scientific Exploitation of Operational Missions–Improved Atmospheric Spectroscopy) spectroscopic information on CO columns from TROPOMI (Tropospheric Monitoring Instrument) shortwave infrared (SWIR) observations was examined. HITRAN 2016 (High Resolution Transmission) and GEISA 2015 (Gestion et Etude des Informations Spectroscopiques Atmosphériques 2015) were used as a reference upon which the spectral fitting residuals, retrieval errors and inferred quantities were assessed. It was found that SEOM–IAS significantly improves the quality of the CO retrieval by reducing the residuals to TROPOMI observations. The magnitude of the impact is dependent on the climatological region and spectroscopic reference used. The difference in the CO columns was found to be rather small, although discrepancies reveal, for selected scenes, in particular, for observations with elevated molecular concentrations. A brief comparison to Total Column Carbon Observing Network (TCCON) and Network for the Detection of Atmospheric Composition Change (NDACC) also demonstrated that both spectroscopies cause similar columns; however, the smaller retrieval errors in the SEOM with Speed-Dependent Rautian and line-Mixing (SDRM) inferred CO turned out to be beneficial in the comparison of post-processed mole fractions with ground-based references.

The proxy method, using the ratio of total column CH4 to CO2 to reduce the effects of common biases, has been used to retrieve column-averaged dry-air mole fraction of CH4 from satellite data. The present study characterizes the remaining scattering effects in the CH4/CO2 ratio component of the Greenhouse gases Observing SATellite (GOSAT) retrieval and uses them for bias correction. The variation of bias between the GOSAT and Total Carbon Column Observing Network (TCCON) ratio component with GOSAT data-derived variables was investigated. Then, it was revealed that the variability of the bias could be reduced by using four variables for the bias correction—namely, airmass, 2 μm band radiance normalized with its noise level, the ratio between the partial column-averaged dry-air mole fraction of CH4 for the lower atmosphere and that for the upper atmosphere, and the difference in surface albedo between the CH4 and CO2 bands. The ratio of partial column CH4 reduced the dependence of bias on the cloud fraction and the difference between hemispheres. In addition to the reduction of bias (from 0.43% to 0%), the precision (standard deviation of the difference between GOSAT and TCCON) was reduced from 0.61% to 0.55% by the correction. The bias and its temporal variation were reduced for each site: the mean and standard deviation of the mean bias for individual seasons were within 0.2% for most of the sites.

Volatile organic compounds (VOCs) are key precursors of particulate matter and tropospheric ozone. Although the terrestrial biosphere is by far the largest source of VOCs into the atmosphere, the emissions of biogenic VOCs remain poorly constrained at regional scale. In this work, we derive top-down biogenic emissions over Europe using weekly-averaged TROPOMI formaldehyde (HCHO) data from 2018 to 2021. The systematic bias of the TROPOMI HCHO columns is characterized and corrected for based on comparisons with FTIR data at seven European stations. The top-down fluxes of biogenic, pyrogenic, and anthropogenic VOC sources are optimized using an inversion framework based on the MAGRITTEv1.1 chemistry transport model and its adjoint. The inversion leads to strongly increased isoprene emissions with respect to the MEGAN-MOHYCAN inventory over the model domain (from 8.1 to 18.5 Tg yr−1 ) which is driven by the high observed TROPOMI HCHO columns in southern Europe. The impact of the inversion on biomass burning VOCs (+13%) and anthropogenic VOCs (−17%) is moderate. An evaluation of the optimized HCHO distribution against ground-based remote sensing (FTIR and MAX-DOAS) and in situ data provides generally improved agreement at stations below about 50° N, but indicates overestimated emissions in northern Scandinavia. Sensitivity inversions show that the top-down emissions are robust with respect to changes in the inversion settings and in the model chemical mechanism. However, the top-down emissions are very sensitive to the bias correction of the observed columns. Furthermore, the use of different a priori emissions has a significant impact on the inversion results due to large differences among bottom-up inventories. In regions with variable meteorology, there is strong week-to-week variability in the observed HCHO columns. The top-down emissions, which are optimized at weekly increments, have a much improved capability of representing these large fluctuations than an inversion using monthly increments. [ABSTRACT FROM AUTHOR]

Recent studies show that deep neural networks (DNNs) are extremely vulnerable to elaborately designed adversarial examples. Adversarial training, which uses adversarial examples as training data, has been proven to be one of the most effective methods of defense against adversarial attacks. However, most existing adversarial training methods use adversarial examples relying on first-order gradients, which perform poorly against second-order adversarial attacks and make it difficult to further improve the robustness of the model. In contrast to first-order gradients, second-order gradients provide a more accurate approximation of the loss landscape relative to natural examples. Therefore, our work focuses on constructing second-order adversarial examples and utilizing them for training DNNs. However, second-order optimization involves computing the Hessian inverse, which typically consumes considerable time. To address this issue, we propose an approximation method that transforms the problem into optimization within the Krylov subspace. Compared with the Euclidean space, the Krylov subspace method typically does not require storing the entire matrix. It only needs to store vectors and intermediate results, avoiding explicitly calculating the complete Hessian matrix. We approximate the adversarial direction by a linear combination of Hessian-vector products in the Krylov subspace to reduce the computation cost. Because of the non-symmetrical Hessian matrix, we use the generalized minimum residual to search for an approximate polynomial solution of the matrix. Our method efficiently reduces computational complexity and accelerates the training process. Extensive experiments conducted on the MNIST, CIFAR-10, and ImageNet-100 datasets demonstrate that our adversarial learning using second-order adversarial samples outperforms other first-order methods, leading to improved model robustness against various attacks.