TROPOMI on board of Sentinel-5 Precursor (S5P) provides continuous daily distribution of several cloud properties, which are required as input for trace-gas retrievals. The operational TROPOMI cloud retrieval is a two-step algorithm. At first, the OCRA (Optical Cloud Recognition Algorithm) computes a radiometric cloud fraction using a broad-band UV/VIS color space approach and later the ROCINN (Retrieval of Cloud Information using Neural Networks) retrieves the cloud height, cloud optical thickness and cloud albedo from NIR measurements in and around the oxygen A-band (~760nm). Within the ROCINN algorithm two different models are possible; the Clouds-as-Reflecting-Boundaries (CRB), where the cloud is a simple Lambertian reflector and the Clouds-as-Layers (CAL), where the cloud is a homogeneous layer of scattering liquid-water spherical particles. There is evidence that some TROPOMI cloud retrievals are contaminated by aerosols. This is particularly true in the following cases: (a) when there is co-existence of clouds and aerosols in the same TROPOMI footprint and (b) when there is a pure aerosol layer, appearing in the TROPOMI cloud product. The latter is usually the case of OCRA deriving an elevated radiometric cloud fraction corresponding to the given aerosol conditions. Then, ROCINN is triggered and returns two additional cloud parameters. Often, the false alarms of elevated OCRA cloud fraction can be identified when ROCINN retrieves a cloud height at the surface level. However, there are cases in which ROCINN cloud outputs do not refer to the surface properties of the scene, but to aerosol layers present in the same TROPOMI footprint. Especially for dust aerosols, which are usually large particles and comparable to the cloud droplet size, we expect more frequently those mixed retrievals. In particular, dust layers with large concentrations (i.e., high aerosol optical depth (AOD)) are better candidates for erroneously retrieved clouds in the TROPOMI L2 product. The TROPOMI aerosol algorithm (TropOMAER) makes use of the L1b reflectances in the UV to derive aerosol information in cloud-free and above-cloud aerosol scenes. With the use of ground-based active and passive remote sensing instruments, we are able to characterize well the vertically resolved cloud and aerosol layers in the lower troposphere. In this work, synergistic ground-based measurements from a PollyXT multiwavelength-Raman-polarization lidar and an AERONET sun-photometer are used to discriminate dust aerosols from clouds in TROPOMI measurements. We have selected ground-based observation sites over which the atmospheric column frequently contains large contributions of desert dust particles.

The knowledge of the surface reflectance is essential for the retrieval of atmospheric trace-gases from satellites. It is required in the conversion of the observed trace gas slant column to the total vertical column by means of a so-called air-mass factor. Although there exists climatological databases based on UV satellite data (e.g. OMI, GOME-2), these have a low spatial resolution and are not appropriate for current and future UV satellite missions like Sentinel-5p/TROPOMI or MTG-S/UVN (Sentinel-4) due their significantly higher spatial and spectral resolution. Current climatologies which are used in operational retrievals provide the Lambertian Equivalent Reflection (LER, e.g. OMI, GOME-2, TROPOMI, see [1,2,3]) and Directional-LER (DLER, e.g. GOME-2, TROPOMI see [3,4]) for selected wavelength in the UV-VIS range and are based on the so-called minimum LER approach, i.e. determine the minimum surface reflectance in the measurement timeframe.We present here a new technique called GE_LER (Geometry-dependent Effective Lambertian Equivalent Reflectivity) based on Machine Learning, which retrieves the DLER from UV satellites in a wavelength range as opposed to the single wavelength approaches of existing climatologies. In this way, dedicated surface reflectivities for specific trace gas retrieval wavelength ranges can be determined. We train a Neural Network with simulated UV spectra, which have been calculated with (V)LIDORT (see [5]). This radiative transfer model is also used for the generation of Air Mass Factors in the operational TROPOMI trace gas retrieval. In this way we reduce the influence of using different radiative transfer models with respect to trace gas retrievals.First results of our GE_LER retrieval for several trace-gases based on TROPOMI data will be shown. References[1] Kleipool (2010), OMI/Aura Surface Reflectance Climatology L3 Global Gridded 0.5 degree x 0.5 degree V3, Greenbelt, MD, USA, Goddard Earth Sciences Data and Information Services Center (GES DISC),, 10.5067/Aura/OMI/DATA3006[2] Tilstra et al. (2017), Surface reflectivity climatologies from UV to NIR determined from Earth observations by GOME-2 and SCIAMACHY, J. Geophys. Res. Atmos. 122, 4084-4111, doi:10.1002/2016JD025940[3] Tilstra et al. (2021), Directionally dependent Lambertian-equivalent reflectivity (DLER) of the Earth's surface measured by the GOME-2 satellite instruments, Atmos. Meas. Tech. 14, 4219-4238, doi:10.5194/amt-14-4219-2021[4] Tilstra et al. (2023), A directional surface reflectance climatology determined from TROPOMI observations, Atmos. Meas. Tech. Discuss. [preprint], doi:10.5194/amt-2023-222, in review[4] Spurr et al. (2008), LIDORT and VLIDORT: Linearized pseudo-spherical scalar and vector discrete ordinate radiative transfer models for use in remote sensing retrieval problems. Light Scattering Reviews, Volume 3, ed. A. Kokhanovsky, Springer

The Geo-Ring for Air Quality consists of three geostationary instruments to monitor the air quality and atmospheric composition over large parts of the northern hemisphere with a high temporal cadence. These are the Korean Geostationary Environmental Monitoring Spectrometer (GEMS, launched 2020), the US-American Tropospheric Emissions: Monitoring of Pollution (TEMPO, launched 2023) and the European UVN spectrometer on Sentinel-4 (S4, to be launched 2025). These geostationary instruments can benefit substantially from the knowledge gained by heritage LEO missions like OMI/Aura, GOME-2/MetOP-ABC and TROPOMI/Sentinel-5P and provide a great synergistic potential to combine the global spatial coverage of the LEO missions with the regional high temporal coverage of the GEO missions.Although trace gases and greenhouse gases are the main focus of the Geo-Ring for Air Quality, knowledge about the presence and characteristics of clouds is a pre-requisite for an accurate retrieval of the aforementioned species for air quality. On top of that, clouds by themselves are an important parameter for climatological studies and applications via their importance and impact on the Earth’s radiation budget.In this contribution, we present the operational cloud product developed for Sentinel-4. It is based on the algorithms called OCRA (Optical Cloud Recognition Algorithm) and ROCINN (Retrieval of Cloud Information using Neural Networks), which are already being in operational use for several heritage missions like GOME-2 and TROPOMI. The main retrieval parameters are cloud fraction, cloud mean height, cloud top height, cloud optical thickness and cloud albedo, achieved via two different cloud models: a simplified Lambertian reflector approach (CRB, clouds as reflecting boundaries) and a physically more realistic scattering layer approach (CAL, clouds as layers). As a testing scenario for the Seninel-4 development, the OCRA algorithm has been adapted to the GEMS instrument. We will show application results and also further comparisons of the S5P OCRA/ROCINN cloud product with the GEMS cloud product.

Owing to its high spatial resolution, the TROPOspheric Monitoring Instrument (TROPOMI) launched in 2017 onboard the Sentinel-5 Precursor (S5P) platform provides important information on global volcanic and anthropogenic SO2 emissions, with an unprecedented level of details.In a recent study (Theys et al., 2021), we proposed an approach called Covariance-Based Retrieval Algorithm (COBRA), different from the classical Differential Optical Absorption Spectroscopy (DOAS). Application of COBRA to TROPOMI SO2 column retrievals leads to a significant reduction of the retrieval noise and biases as compared to the TROPOMI operational (DOAS-based) SO2 product. COBRA even reveals new emission sources in long-term averaged SO2 maps.In view of a future operational deployment (planned end 2024), the COBRA SO2 scheme is being implemented as part of the Copernicus S5-P Product Algorithm Laboratory (PAL). In this poster, we give an update of TROPOMI COBRA SO2 results. The latest developments of COBRA S5P-PAL v2 algorithm are presented and discussed. The pre-operational S5P-PAL environment enables a full reprocessing of TROPOMI data. For several examples, we illustrate the COBRA data set for the long-term monitoring of SO2 columns over both anthropogenic and volcanic scenes. Finally, possible future developments of COBRA are discussed. Theys, N., Fioletov, V., Li, C., De Smedt, I., Lerot, C., McLinden, C., Krotkov, N., Griffin, D., Clarisse, L., Hedelt, P., Loyola, D., Wagner, T., Kumar, V., Innes, A., Ribas, R., Hendrick, F., Vlietinck, J., Brenot, H., and Van Roozendael, M.: A Sulfur Dioxide Covariance-Based Retrieval Algorithm (COBRA): application to TROPOMI reveals new emission sources, Atmos. Chem. Phys., 21, 16727–16744, https://doi.org/10.5194/acp-21-16727-2021, 2021.

Due to their fast computational performance and accuracy, neural networks are nowadays commonly used in the context of remote sensing. The issue of performance is especially important in the context of big data and near-real-time (NRT) operational processing. Classical retrieval algorithms typically use a radiative transfer model (RTM) as a forward model to solve the inverse problem of inferring the quantities of interest from the measured spectra. However, these RTMs are often computationally very expensive and therefore replacing them by a NN is desirable to increase performance. But the application of NNs is not straightforward and there are at least two main approaches: For the inverse problem in atmospheric remote sensing, there are two main approaches to apply NNs: 1. NNs used as forward model, where a NN accurately approximates the radiative transfer model and can thus replace it in the inversion algorithm 2. NNs for solving the inverse problem, where a NN is trained to infer the atmospheric parameters from the measurement directly The first approach is more straightforward to apply. However, the inversion algorithm still faces many challenges, as the spectral fitting problem is generally ill-posed. Therefore, local minima are possible and the results often depend on the selection of the a-priori values for the retrieval parameters. For the second case, some of these issues can be avoided: no a-priori values are necessary, and as the training of the NN is performed globally, i.e. for many training samples at once, this approach is potentially less affected by local minima. However, due to the black-box nature of a NN, no indication about the quality of the results is available. In order to address this issue, novel methods like Bayesian neural networks (BNNs) or invertible neural networks (INNs) have been presented in recent years. This allows the characterization of the retrieved values by an estimate of uncertainty describing a range of values that are probable to produce the observed measurement. We apply and evaluate both approaches for the retrieval of cloud properties and consider their potential as operational algorithms for current (Sentinel-5P) and future (Sentinel-4) Copernicus atmospheric composition missions.

The retrieval of cloud parameters from the atmospheric Sentinel missions requires Earth reflectance measurements from a set of spectral bands. The ground pixels of the involved spectral bands should be fully aligned, but when they are not, special treatment is required within the operational algorithms. This so-called inter-band spatial misregistration of passive spectrometers is present when the Earth reflectance measurements in different spectral bands are captured by different spectrometers. The cloud retrieval algorithm requires reflectance measurements in the UV(ultraviolet)-VIS (visible) band, where the first cloud parameter (i.e., radiometric cloud fraction) is retrieved from OCRA (Optical Cloud Recognition Algorithm). In addition, Earth reflectances in the NIR (near-infrared) band are needed for the retrieval of two additional cloud parameters (i.e., cloud height and cloud albedo or cloud-top height and optical thickness) from the ROCINN (Retrieval of Cloud Information using Neural Networks) algorithm. In the former TROPOMI (TROPOspheric Monitoring Instrument)/S5P (Sentinel-5 Precursor) retrieval, a coregistration scheme of the derived cloud parameters from the source band to the target band based on pre-calculated mapping weights from UV-VIS to NIR and vice versa is applied. In this paper we present a new scheme for the coregistration of the TROPOMI cloud parameters using collocated VIIRS (Visible Infrared Imaging Radiometer Suite)/SNPP (Suomi National Polar-orbiting Partnership) information. The new coregistration scheme based on the VIIRS data improves the TROPOMI cloud product quality and allows the addition of cloud information for the first (westernmost) TROPOMI UVIS ground pixel. In practice, the latter means that a significant number of valid data points are included in the TROPOMI cloud, total ozone, SO2 and HCHO product since 26 November 2023 (orbit 31705), when the new coregistration scheme became operational. From a comparison analysis between the two techniques, we found that the largest differences mainly appear for inhomogeneous scenes. From a validation exercise of TROPOMI against VIIRS in the across-track direction, we found that the old coregistration scheme tends to smooth out cloud structures along the scan line, whereas such structures can be maintained with the new scheme. The need to implement a similar inter-band spatial coregistration scheme is foreseen for the Sentinel-4/MTG-S (Meteosat Third Generation - Sounder) and Sentinel-5/MetOp-SG (Meteorological Operational Satellite - Second Generation) missions. In the case of the Sentinel-4 instrument, the external cloud information will originate from collocated data captured by the FCI (Flexible Combined Imager) on board the MTG-I (Meteosat Third Generation - Imager) satellite.

In recent decades, there has been an increasing interest in making use of satellite measurements for identifying trends in atmospheric composition and climate. Instruments like GOME-2 and TROPOMI are dedicated to air-quality and global trace gas monitoring. For the accurate retrieval of columnar information of the trace gases, cloud correction is necessary. This work is meant to examine the quality of the GOME-2 operational cloud product from AC SAF and to propose enhancements of the current dataset to improve the retrieval of the NO2 and HCHO tropospheric gases.

The retrieval of atmospheric composition from space-based measurements, by e.g., Sentinel-5p TROPOMI, is strongly affected by radiative interferences with clouds. Dedicated cloud data products, typically retrieved from measurements by the same sounder, are therefore essential. Cloud information is used to filter data and as input to the modelling of atmospheric radiative transfer and the conversion of slant column densities into vertical column densities. The three main TROPOMI cloud retrieval algorithms are: (i) L2_CLOUD OCRA/ROCINN CAL (Optical Cloud Recognition Algorithm/Retrieval of Cloud Information using Neural Networks; Clouds-As-Layers), (ii) L2_CLOUD OCRA/ROCINN CRB (Clouds-as Reflecting Boundaries), and (iii) the S5P support product FRESCO-S (Fast Retrieval Scheme for Clouds from Oxygen absorption bands for Sentinel). The cloud variables provided by these products (radiometric cloud fraction, cloud (top) height, and cloud albedo/cloud optical thickness) are subsequently used in the retrieval of the TROPOMI trace gas products. The quality of cloud products and trace gas products is routinely assessed by the ESA/Copernicus Atmospheric Mission Performance Cluster (ATM-MPC) validation service, with ad hoc support from Sentinel-5p Validation Team (S5PVT) AO projects. Version upgrades have had a significant impact on the characteristics of S5P cloud data. The change of the wavelength window in the FRESCO product since version 1.4 (‘FRESCO-wide’) leads to a clear increase in the height of low clouds with a large impact on the tropospheric NO2 retrieval (van Geffen, 2022), and improving the validation results regarding the tropospheric and total NO2 column. The first upgrades of the ROCINN products (from v1 to v2.1-v2.3) led to an increase in correlation with CLOUDNET cloud height, but to a more negative bias for the low clouds, with ROCINN CRB cloud height even dropping below the CLOUDNET cloud base height on average. However, this effect seems alleviated with the latest upgrade to v2.4. The impact on the HCHO validation results is investigated but is less clear compared to the NO2 case. To resolve the discontinuities due to the processor version jumps, a full mission reprocessing is currently ongoing and largely carried out for the L2_CLOUD and FRESCO-S products. The reprocessed ROCINN data have a lower dispersion and higher correlation with respect to the CLOUDNET cloud heights. The bias of the L2_CLOUD OCRA/ROCINN CAL CTH becomes more negative, but that of L2_CLOUD OCRA/ROCINN CRB CH bias improves. Finally, we also discuss the impact of the FRESCO-S reprocessing on the validation results.

<p>Due to their fast computational performance, neural networks (NNs) are nowadays commonly used in the context of remote sensing. The issue of performance is especially important in the context of big data and operational processing. Classical retrieval algorithms often use a radiative transfer model (RTM) as forward model with which an optimization algorithm can then solve the inverse problem of inferring the quantities of interest from the measured spectra. However, these RTMs are usually computationally very expensive and therefore replacing them by a NN is desirable to increase performance. But the application of NNs is not straightforward and there are at least two main approaches:</p> <p>1. NNs used as forward model, where a NN approximates the radiative transfer model and can thus replace it in the inversion algorithm</p> <p>2. NNs for solving the inverse problem, where a NN is trained to infer the atmospheric parameters from the measurement directly</p> <p>The first approach is more straightforward to apply. However, the inversion algorithm still faces many challenges, as the spectral fitting problem is generally ill-posed. Therefore, local minima are possible and the results often depend on the selection of the a-priori values for the retrieval parameters.</p> <p>For the second case, some of these issues can be avoided: no a-priori values are necessary, and as the training of the NN is performed globally, i.e. for many training samples at once, this approach is potentially less affected by local minima. However, due to the black-box nature of a NN, no indication about the quality of the results is available. In order to address this issue, novel methods like Bayesian neural networks (BNNs) or invertible neural networks (INNs) have been presented in recent years. This allows the characterization of the retrieved values by an estimate of uncertainty describing a range of values that are probable to produce the observed measurement. We apply and evaluate these new BNN and INN methods for the retrieval of cloud properties from TROPOMI in order to demonstrate their potential as operational algorithms for current (Sentinel-5P) and future (Sentinel-4 and Sentinel-5) Copernicus atmospheric composition missions.</p>

The status and most recent developments of the operational L2 Cloud product will be presented in this contribution for the ongoing Sentinel-5 Precursor and upcoming Sentinel-4 missions. These Copernicus missions are focused on atmospheric composition, operate in the UV/VIS/NIR/(SWIR) spectral region and comprise the retrieval of trace gases, greenhouse gases, aerosol and cloud properties. A good knowledge about the latter, i.e. the presence and characteristics of clouds, is a pre-requisite for an accurate retrieval of the aforementioned trace gases and greenhouse gases. Additionally, clouds are by themselves an interesting indicator to measure and monitor because of their contribution to the radiation budget, and hence, impact on climatological applications. The algorithms for retrieving the operational cloud products from TROPOMI onboard Sentinel-5 Precursor and the UVN spectrometer onboard Sentinel-4 are called OCRA (Optical Cloud Recognition Algorithm) and ROCINN (Retrieval of Cloud Information using Neural Networks) and both have their heritage with GOME/ERS-2 and GOME-2 MetOp-A/B/C, where they have already been successfully implemented in an operational environment. OCRA applies a broad band color space approach to the measured reflectance in order to retrieve a radiometric cloud fraction that is used as an a priori input to ROCINN, which retrieves cloud top height, cloud optical thickness and cloud albedo from measurements of sun-normalized radiances in the NIR in and around the oxygen A-band. The cloud parameters retrieved by ROCINN are provided for two different cloud models. The Clouds-as-Layers (CAL) model treats clouds as layers of scattering water droplets, which is physically more realistic than the second model, Clouds-as-Reflecting Boundaries (CRB), which treats a cloud as a simple Lambertian reflector. In addition, this contribution will cover initial results of applying the OCRA algorithm to the recently launched Korean geostationary GEMS instrument. Applying the algorithm to GEMS does provide a great opportunity to test the performance under a geostationary configuration and to transfer lessons learned in a synergetic way to the Sentinel-4 development. Also, initial comparisons of the S5P and GEMS cloud products will be shown.

In remote sensing, the quantities of interest (e.g. the composition of the atmosphere) are usually not directly observable but can only be inferred indirectly via the measured spectra. To solve these inverse problems, retrieval algorithms are applied that usually depend on complex physical models, so-called radiative transfer models (RTMs). RTMs are very accurate, however also computationally very expensive and therefore often not feasible in combination with the strict time requirements of operational processing of satellite measurements. With the advances in machine learning, the methods of this field, especially deep neural networks (DNN), have become very promising for accelerating and improving the classical remote sensing retrieval algorithms. However, their application is not straightforward but instead quite challenging as there are many aspects to consider and parameters to optimize in order to achieve satisfying results. In this presentation we show a general framework for replacing the RTM, used in an inversion algorithm, with a DNN that offers sufficient accuracy while at the same time increases the processing performance by several orders of magnitude. The different steps, sampling and generation of the training data, the selection of the DNN hyperparameters, the training and finally the integration of the DNN into an operational environment are explained in detail. We will also focus on optimizing the efficiency of each step: optimizing the generation of training samples through smart sampling techniques, accelerating the training data generation through parallelization and other optimizations of the RTM, application of tools for the DNN hyperparameter optimization as well as the use of automation tools (source code generation) and appropriate interfaces for the efficient integration in operational processing systems. This procedure has been continuously developed throughout the last years and as a use case, it will be shown how it has been applied in the operational retrieval of cloud properties for the Copernicus satellite sensors Sentinel-4 (S4) and TROPOMI/Sentinel-5 Precursor (S5P).

Atmospheric formaldehyde (HCHO) is a secondary product in the destruction of non-methane volatile organic compounds (NMVOCs), through both natural and anthropogenic processes. With a relatively short lifetime of a few hours, the HCHO concentrations are usually localised close to their source. Measuring HCHO from space is therefore highly relevant in obtaining information on NMVOC emissions and their role in air quality and climate. HCHO retrievals from space have so far been limited to polar orbiting sensors with a fixed local overpass time. The Geostationary Environment Monitoring Spectrometer (GEMS), launched on-board the GEO-KOMPSAT-2B satellite in February 2020 is the first geostationary sensor dedicated to air quality and atmospheric composition measurements. GEMS (observing South-East Asia hourly) will be complemented by TEMPO in 2022 (United States) and Sentinel-4 in 2023 (Europe and Northern Africa). Those instruments will provide an unprecedented hourly revisit time in their respective spatial domains. However, geostationary sensors make fundamentally different demands on the HCHO algorithm as compared to polar sensors. In this work, we present DOAS tropospheric column retrieval results for HCHO from GEMS. In order to fit the SCD, a precise wavelength calibration is applied and potential changes in the instrumental line shape are accounted for. Polarisation spectral structures and scene heterogeneity effects are included, and a background correction and destriping procedure dedicated to geostationary observations is also developed. Air mass factors are calculated using auxiliary data consistent with the TROPOMI operational product. We compare our first results with those from TROPOMI in the early afternoon and with the GEMS HCHO operational product. Finally, we examine the diurnal variations observed with GEMS over different emission sources. MAX-DOAS measurements are used to validate and interpret the observed hourly variations.

The study of ice clouds properties is of central importance to further understand the role of ice clouds in climate system processes. Therefore, it is crucial to perform accurate ice cloud retrievals in satellite-based systems in order to provide reliable information about the cloud microphysical, macrophysical and optical properties. Current and future satellite missions like Sentinel-5 Precursor (S5P), Sentinel-4 (S4), and Sentinel-5 (S5) are designed to monitor the air quality and greenhouse gases. The cloud retrieval algorithm used operationally for these missions is ROCINN (Retrieval Of Cloud Information using Neural Networks) which retrieves the cloud top height (CTH), cloud optical depth (COD) and cloud albedo (CA) from measurements in the NIR in the O2 A-band (755-771 nm). ROCINN considers two cloud models: Clouds as Reflecting Boundaries (CRB) and Clouds As scattering Layers (CAL). In this work we will present the latest developments including the ice cloud retrieval performed using the VLIDORT radiative transfer (RT) model containing ice cloud parametrization. This study investigates the performance of ROCINN for ice cloud retrieval for several test scenarios adapted from Level 2 operational data. The selected datasets contain partially and fully cloudy scenarios for ice clouds placed at different CTH and for different COD. The retrieved CTH and COD for ice clouds are evaluated for the TROPOMI/S5P and S4 satellites.

The aim of this paper is to highlight how TROPOspheric Monitoring Instrument (TROPOMI) trace gas data can best be used and interpreted to understand event-based impacts on air quality from regional to city scales around the globe. For this study, we present the observed changes in the atmospheric column amounts of five trace gases (NO2, SO2, CO, HCHO, and CHOCHO) detected by the Sentinel-5P TROPOMI instrument and driven by reductions in anthropogenic emissions due to COVID-19 lockdown measures in 2020. We report clear COVID-19-related decreases in TROPOMI NO2 column amounts on all continents. For megacities, reductions in column amounts of tropospheric NO2 range between 14 % and 63 %. For China and India, supported by NO2 observations, where the primary source of anthropogenic SO2 is coal-fired power generation, we were able to detect sector-specific emission changes using the SO2 data. For HCHO and CHOCHO, we consistently observe anthropogenic changes in 2-week-averaged column amounts over China and India during the early phases of the lockdown periods. That these variations over such a short timescale are detectable from space is due to the high resolution and improved sensitivity of the TROPOMI instrument. For CO, we observe a small reduction over China, which is in concert with the other trace gas reductions observed during lockdown; however, large interannual differences prevent firm conclusions from being drawn. The joint analysis of COVID-19-lockdown-driven reductions in satellite-observed trace gas column amounts using the latest operational and scientific retrieval techniques for five species concomitantly is unprecedented. However, the meteorologically and seasonally driven variability of the five trace gases does not allow for drawing fully quantitative conclusions on the reduction in anthropogenic emissions based on TROPOMI observations alone. We anticipate that in future the combined use of inverse modeling techniques with the high spatial resolution data from S5P/TROPOMI for all observed trace gases presented here will yield a significantly improved sector-specific, space-based analysis of the impact of COVID-19 lockdown measures as compared to other existing satellite observations. Such analyses will further enhance the scientific impact and societal relevance of the TROPOMI mission.

Abstract. We present a new tropospheric ozone data set based on TROMOMI/Sentinel-5 Precursor (S5P) total ozone mea- surements combined stratospheric ozone data from the Belgian Assimilation System for Chemical ObsErvations (BASCOE). BASCOE is constrained by assimilating ozone observations from the microwave limb sounder (MLS). The tropospheric ozone algorithm is similar to the well established OMI-MLS or OMPS-MERRA-2 retrieval. Compared to this we gain spatial resolu- tion when applying the algorithm to TROPOMI data (5.5 × 3.5 km2) and extend these data record into the future. Compared to the OMPS-MERRA-2 data a mean positive bias of ≈ 3 DU is found. A small negative bias of about 2 DU is observed in the tropics relative to the operational TROPOMI tropospheric (S5P_O3_TCL) data based on the CCD algorithm. The new tropospheric ozone data (S5P-BASCOE) is compared to a set of globally distributed ozone sondes data integrated up to the tropopause level. For the comparison both the mean of the satellite observations around the sounding station and the closest column data are used. Depending on the latitude the S5P-BASCOE deviate from the sondes and between 0 and 5 DU, indicating a good agreement. However, some exceptional larger positive deviation up to ±10 DU are found. The monthly mean tropospheric column as well as time series for selected places showed the expected spatial and temporal pattern.

Ecosystems and human health are severely harmed by elevated concentrations of tropospheric ozone, in the short and the long term. Monitoring ozone at all relevant spatial and temporal scales simultaneously is a challenge for a global observing system due to the large variability of ozone levels in the troposphere. Space-based sensors provide near-global coverage at the synoptic scale, but their accuracy is limited since the large stratospheric O3 column shields the view on the relatively small tropospheric ozone concentrations. In contrast, in-situ soundings by balloons are sparse, but these are more accurate and at a high vertical resolution. As a result, the geophysical information that can be inferred from tropospheric ozone data records differs. We present a comprehensive comparison of the spatial and temporal patterns in tropical tropospheric ozone column observations by nadir-viewing satellite sensors (Sentinel-5 Precursor/TROPOMI, EOS-Aura/OMI and Metop-B/GOME-2) and ozonesondes for the period 2018-2020. We discuss how each data record perceives well-known structures and cycles such as the zonal wave-one, the seasonal cycle and biomass burning periods. Imprints of (sensor-dependent) sampling characteristics are generally less relevant on large scales. However, these can dominate the uncertainty budget when satellite data are used at their finest sampling resolution. Nonetheless, we recognise the signature of the Madden-Julian Oscillation and hints of Kelvin wave activity.

Space-born atmospheric composition measurements, like those from Sentinel-5p TROPOMI, are strongly affected by the presence of clouds. Dedicated cloud data products, typically retrieved with the same sensor, are therefore an important tool for the provider of atmospheric trace gas retrievals. Cloud products are used for filtering and modification of the modelled radiative transfer. In this work, we assess the quality of the cloud data derived from Copernicus Sentinel-5 Precursor TROPOMI radiance measurements. Three cloud products are considered: (i) L2_CLOUD OCRA/ROCINN CAL (Optical Cloud Recognition Algorithm/Retrieval of Cloud Information using Neural Networks; Clouds-As-Layers), (ii) L2_CLOUD OCRA/ROCINN CRB (same; Clouds-as Reflecting Boundaries), and (iii) the S5p support product FRESCO-S (Fast Retrieval Scheme for Clouds from Oxygen absorption bands for Sentinel). These cloud products are used in the retrieval of several S5p trace gas products (e.g., ozone columns and profile, total and tropospheric nitrogen dioxide, sulfur dioxide, formaldehyde). The quality assessment of these cloud products is carried out within the framework of ESA’s Sentinel-5p Mission Performance Centre (MPC) with support from AO validation projects focusing on the respective atmospheric gases. Cloud height data from the three S5p cloud products is compared to radar/lidar based cloud profile information from the ground-based networks CLOUDNET and ARM. The cloud height from S5p CLOUD CRB and S5p FRESCO are on average 0.6 km below the cloud mid-height of CLOUDNET measurements, and the cloud top height from S5p CLOUD CAL is on average 1 km below CLOUDNET’s cloud top height. However, the comparison is different for low and high clouds, with S5p CLOUD CAL cloud top height being only 0.3 km below CLOUDNET’s for low clouds. The radiometric cloud fraction and cloud (top) height are compared to those of other satellite cloud products like Aura OMI O2-O2. While the latitudinal variation is often similar, offsets are encountered. Recently, major S5p cloud product upgrades were released for S5p OCRA/ROCINN (July 2020) and for S5p FRESCO (December 2020), leading to a decrease of the ROCINN CRB cloud height and an increase of the FRESCO cloud height on average. Moreover, a major change in the ROCINN surface albedo treatment leads to a clear improvement of the comparison with CLOUDNET at the complicated sea/land/ice/snow site Ny-Alesund.

Ozone data derived from the Tropospheric Monitoring Instrument (TROPOMI) sensor on board the Sentinel-5 Precursor satellite show exceptionally low total ozone columns in the polar region of the Northern Hemisphere (Arctic) in spring 2020. Minimum total ozone column values around or below 220 Dobson units (DU) were seen over the Arctic for 5 weeks in March and early April 2020. Usually the persistence of such low total ozone column values in spring is only observed in the polar Southern Hemisphere (Antarctic) and not over the Arctic. These record low total ozone columns were caused by a particularly strong polar vortex in the stratosphere with a persistent cold stratosphere at higher latitudes, a prerequisite for ozone depletion through heterogeneous chemistry. Based on the ERA5, which is the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis, the Northern Hemisphere winter 2019/2020 (from December to March) showed minimum polar cap temperatures consistently below 195 K around 20 km altitude, which enabled enhanced formation of polar stratospheric clouds. The special situation in spring 2020 is compared and discussed in context with two other Northern Hemisphere spring seasons, namely those in 1997 and 2011, which also displayed relatively low total ozone column values. However, during these years, total ozone columns below 220 DU over several consecutive days were not observed in spring. The similarities and differences of the atmospheric conditions of these three events and possible explanations for the observed features are presented and discussed. It becomes apparent that the monthly mean of the minimum total ozone column value for March 2020 (221 DU) was clearly below the respective values found in March 1997 (267 DU) and 2011 (252 DU), which highlights the special evolution of the polar stratospheric ozone layer in the Northern Hemisphere in spring 2020. A comparison with a typical ozone hole over the Antarctic (e.g., in 2016) indicates that although the Arctic spring 2020 situation is remarkable, with total ozone column values around or below 220 DU observed over a considerable area (up to 0.9 million km2), the Antarctic ozone hole shows total ozone columns typically below 150 DU over a much larger area (of the order of 20 million km2). Furthermore, total ozone columns below 220 DU are typically observed over the Antarctic for about 4 months.

Accurate knowledge of cloud properties is essential to the measurement of atmospheric composition from space. In this work we assess the quality of the cloud data from three Copernicus Sentinel-5 Precursor (S5P) TROPOMI cloud products: (i) S5P OCRA/ROCINN_CAL (Optical Cloud Recognition Algorithm/Retrieval of Cloud Information using Neural Networks;Clouds-As-Layers), (ii) S5P OCRA/ROCINN_CRB (Clouds-as-Reflecting Boundaries), and (iii) S5P FRESCO-S (Fast Retrieval Scheme for Clouds from Oxygen absorption bands – Sentinel). Target properties of this work are cloud-top height and cloud optical thickness (OCRA/ROCINN_CAL), cloud height (OCRA/ROCINN_CRB and FRESCO-S), and radiometric cloud fraction (all three algorithms). The analysis combines (i) the examination of cloud maps for artificial geographical patterns, (ii) the comparison to other satellite cloud data (MODIS, NPP-VIIRS, and OMI O2–O2), and (iii) ground-based validation with respect to correlative observations (30 April 2018 to 27 February 2020) from the Cloudnet network of ceilometers, lidars, and radars. Zonal mean latitudinal variation of S5P cloud properties is similar to that of other satellite data. S5P OCRA/ROCINN_CAL agrees well with NPP VIIRS cloud-top height and cloud optical thickness and with Cloudnet cloud-top height, especially for the low (mostly liquid) clouds. For the high clouds, S5P OCRA/ROCINN_CAL cloud-top height is below the cloud-top height of VIIRS and of Cloudnet, while its cloud optical thickness is higher than that of VIIRS. S5P OCRA/ROCINN_CRB and S5P FRESCO cloud height are well below the Cloudnet cloud mean height for the low clouds but match on average better with the Cloudnet cloud mean height for the higher clouds. As opposed to S5P OCRA/ROCINN_CRB and S5P FRESCO, S5P OCRA/ROCINN_CAL is well able to match the lowest CTH mode of the Cloudnet observations. Peculiar geographical patterns are identified in the cloud products and will be mitigated in future releases of the cloud data products.

Sentinel-5 Precursor and Sentinel-4 are atmospheric Copernicus missions focused on trace gas, greenhouse gas, aerosol and cloud retrieval and operate in the UV/VIS/NIR/(SWIR) spectral region. A key ingredient for the retrieval of the aforementioned trace gases and greenhouse gases is a precise knowledge of the presence of clouds. On top of that, clouds are by themselves interesting to measure and monitor because of their contribution to the radiation budget, and hence, impact on climatological applications. In this contribution, we present the algorithms for retrieving the operational cloud products from TROPOMI onboard Sentinel-5 Precursor and the UVN spectrometer onboard Sentinel-4. These are called OCRA (Optical Cloud Recognition Algorithm) and ROCINN (Retrieval of Cloud Information using Neural Networks) and both have their heritage with GOME/ERS-2 and GOME-2 MetOp-A/B/C, where they have already been successfully implemented in an operational environment. OCRA employs a broad band color space approach to determine a radiometric cloud fraction and ROCINN retrieves cloud top height, cloud optical thickness and cloud albedo from NIR measurements in and around the oxygen A-band, taking as a priori input the cloud fraction computed by OCRA. The cloud parameters retrieved by ROCINN are provided for two different cloud models. One which treats clouds more realistically as layers of scattering water droplets (clouds-as-layers, CAL) and one which treats clouds as simple Lambertian reflectors (clouds-as-reflecting boundaries, CRB). The current status of the algorithms is presented along with recent developments and improvements.