We present new products of global atmospheric aerosol from ESA missions intended to provide climate-quality long term datasets, and a near real time (NRT) product for assimilation in meteorological, atmospheric transport and air quality models from the Copernicus Sentinel-3 mission. The parameters retrieved include aerosol optical depth (AOD), fine mode fraction, dust AOD, Angstrom exponent and retrieval uncertainty. The near real time product makes use of the SLSTR dual view over land, and full swath retrieval over ocean. In addition, a second non-time critical (NTC) product makes synergistic use of the two optical sensors, OLCI and SLSTR to provide potential for greater coverage and retrieval accuracy. Aerosol products are retrieved at a 4.5km resolution for Sentinel-3 products, and 9km for (A)ATSR series.

The method for global aerosol retrieval using SLSTR is based on algorithms developed under the ESA Aerosol Climate Change Initiative (CCI), to develop and evaluate retrieval of aerosol properties and their uncertainties. This has led to generation of the first 17-year dataset from global aerosol retrieval from ERS-2 and ENVISAT (1995-2012), and evaluated using the global AERONET sun photometer network. Validation of the (A)ATSR record shows good correspondence when compared against global AERONET measurements, with correlation coefficient R= ~0.8 over land, and 0.9 over ocean, and shows low bias (globally 0.01-0.02 in AOD at 550nm). Uncertainties are derived by propagation of estimated instrument and model errors within the retrieval, and show good agreement with distribution of measured errors in validation. Retrieval accuracy is greatest over darker surfaces, and additionally the spatial pattern of errors differs between Sentinel-3 and (A)ATSR due to differences in instrument pointing. Recent research has aimed to improve retrieval over bright desert surfaces, and establish continuity between the instrument records. A version of the (A)ATSR record has been accepted under the Observations for Model Intercomparisons Project (OBS4MIPS) to provide datasets suitable for climate model intercomparison, and also forms a contribution to the Copernicus Climate Change Service (C3S). Recent improvements in the long-term archive, and first results of testing the new products for Sentinel-3 will be presented.

Team: G. de Leeuw, O. Dubovik, M. Garay, A. Heckel, H.T. Jethva, R. Kahn, S. Kinne, M. Kosmale, P. North, T. Popp, L. Remer, M. Schulz, A. Sayer, R. Levy, P. Lytvynov, A. Lyapustin, O. Torres et al.
During the last four decades, numerous satellite sensors were launched and algorithms have been generated to study aerosols. However, the lifetime of the sensors is limited and datasets from different sensors and algorithms should be merged to study the long-term aerosol optical depth (AOD) trends for several decades.
The exercise to study the possibility to build an AOD long-term data record has been initiated by the 5th AeroSat in Helsinki, 2017.
First, AOD datasets (publicly available official products) from different satellites and algorithms (Table 1) were collected. Seasonal and yearly aggregated AODs were calculated from the monthly means. AOD time series were visualized for different locations: global over land, global over ocean, mainland China.
For the AOD inter-comparison, two reference years, 2008 and 2018, were chosen. For the reference years, the AOD difference maps were plotted, taking as a “reference” MODIS/Terra, which has better coverage over land and ocean and longer data record. AOD data from the afternoon satellite missions were also compared with MODIS/Aqua AOD.
AOD monthly aggregates from satellites were compared with the AERONET monthly AOD. The analysis was performed for all available collocations and separately for each AERONET station.
The L2 daily AOD inter-comparison for the reference years is planned.
The merging technique will be discussed based on the results from the L2/L3 AOD products inter-comparison, when the last is completed.

Table 1. Instruments/algorithms participation in the inter-comparison exercise

The CISAR (Combined Inversion of Surface and Atmosphere pRoperties) algorithm is exploited in the framework of the ESA-SEOM CIRCAS (ConsIstent Retrieval of Cloud Aerosol Surface) project, aiming at providing a set of atmospheric (cloud and aerosol) and surface reflectance products derived from S3A/SLSTR observations using the same radiative transfer physics and assumptions. CISAR is an advance algorithm developed by Rayference previously deployed for the retrieval of aerosol single scattering properties and surface reflectance from satellite observations (both geostationary and polar orbiting satellites), based on the inversion of a fast radiative transfer model (FASTRE). The aerosol properties are retrieved with continuous variation in the solution space. The latter is defined by three aerosol vertices with different single scattering properties. In order to extend the retrieval with CISAR to clouds, cloud vertices (water and ice clouds) are also considered. The aerosol (cloud) properties are retrieved as a combination of the aerosol (cloud) vertices.

Traditionally, different approaches are exploited to retrieve the different Earth system components, which could lead to inconsistent data sets. The simultaneous retrieval of different atmospheric and surface variables over any type of surface (including bright surfaces and water bodies) with the same forward model and inversion scheme ensures the consistency among the retrieved Earth system components. Additionally, pixels located in the transition zone between pure clouds and pure aerosols are often discarded from both cloud and aerosol algorithms. This “twilight zone” can cover up to 30% of the globe. A consistent retrieval of both cloud and aerosol single scattering properties with the same algorithm could help filling this gap. The CIRCAS project ultimately aims at overcoming the need of an external cloud mask, letting the CISAR algorithm discriminate between aerosol and cloud properties. This would also help reducing the overestimation of aerosol optical thickness in cloud contaminated pixels. The surface reflectance product is delivered both for cloud-free and cloudy observations.

The adaptation of the CISAR algorithm for the cloud single scattering properties retrievals will be described. The ongoing CIRCAS results will be presented both in terms of timeseries and images. The CISAR retrieval will be evaluated against independent datasets.

In the inversion of Essential Climate Variables from satellite-based observations uncertainties of all input data and assumptions propagate through the different levels of processing, starting from level1 (calibrated reflectances per pixel) via level2 (retrieved geophysical parameters per pixel) to level3 (gridded daily and monthly averages of all pixels values within the grid cell). Thorough estimation of those uncertainties and rigid propagation through the processing levels are a challenging task.

In the ESA Climate Change Initiative (CCI) project on aerosols (Aerosol_cci) pixel-level (level2) uncertainties for total AOD through uncertainty propagation of average level1 reflectance uncertainties were developed, validated and improved. Accordingly the products contain for each observed pixel a value of its retrieved AOD and its associated uncertainty. Those uncertainties can typically not yet consider or validate all relevant uncertainty contributions (e.g. cloud-mask induced uncertainties). These pixel-level uncertainties fulfill the key requirement for data assimilation of the level2 products into atmospheric models. Propagation of level2 uncertainties to gridded level3 products was limited to average level2 uncertainties per grid cell, which does not properly take into account any of their correlation structures.

In the EU Horizon2020 project “Fidelity and uncertainty in climate data records from Earth Observations” (FIDUCEO) uncertainty propagation through different processing levels is elaborated further. This starts from characterizing in more detail the uncertainties of the input measured reflectances with random (uncorrelated), structured (correlated with spatial / temporal range) and common (globally correlated) elements. These separate uncertainty contributions can then be propagated further to level2 and level3 products while properly considering their different correlation structures. In addition, assumptions (e.g. model-based aerosol type climatology) and auxiliary information (surface brightness estimated from a vegetation index and a mid-infrared channel) used in the retrieval inversion are also analysed in terms of their uncertainties and correlation structures. All this is documented in an analysis tree which identifies for each element of the measurement operator its inputs and characterizes their uncertainties and correlation structures in separate effects tables. Consistent propagation of uncertainties to the gridded daily and monthly level3 products can then be achieved by separately propagating elements with different correlation structures.

In this paper, after a brief discussion of achieved aerosol uncertainty propagation in CCI, a very simple example for Aerosol Optical Depth inversion from AVHRR over land will be analysed and the propagation through the different levels will be demonstrated. Analysis trees for level2 and level3 products and underlying effects tables for this AOD product developed in FIDUCEO will be discussed and example processing results of different uncertainty contributions at different levels will be shown and discussed.

The figure shows in the top the level2 (inversion) measurement-function centered analysis tree for AVHRR AOD and in the bottom, as one example the effects table for the aerosol type dependence in the inversion.

The availability of high (temporal, spatial) resolution aerosol time series is of crucial importance for the development of climate modelling applications. Such climate data records must cover radiative properties (e.g. the extinction at several wavelengths) and microphysical properties (e.g. size information).
It was the main objective of the ESA Climate Change Initiative (CCI, 2010-2017), and in particular of the project Aerosol_CCI devoted to both tropospheric and stratospheric aerosols, to contribute significantly to the development of such kind of climate data records (CDRs) and to support the algorithm development needed for the obtention of high-quality products. Nowadays, the development of CDRs continues, and the products developed in Aerosol_CCI and their successors are delivered to the Copernicus Climate Change services (C3S) to ensure their distribution to the scientific community.
In this work, we focus on stratospheric aerosols properties retrieved from GOMOS, an experiment based on the technique of stellar occultation that flew onboard ENVISAT and observed the atmosphere on a global scale during about 10 years (2002-2012). We use here the AerGOM algorithm that was specifically developed to retrieve aerosol properties, including the spectral dependence of their radiative properties, over the UV-visible-near IR range. From the retrieved quantities, we derived time series with a 5-day time resolution and 3D spatial resolution of 5° latitude x 60° longitude x 1 km altitude.
In this presentation, we’ll give an overview of the most recent developments of this work, including the latest improvements of the different stratospheric aerosols CDRs, and mainly of the particle size distribution retrieval. Beyond the retrieval aspects, we will present a selection of results illustrating the performances of the CDRs and comparisons with other datasets. Using simulations of the global chemistry-climate model EMAC (general circulation model ECHAM5 coupled with the MESSy Atmospheric Chemistry module), we will analyse the size information provided by GOMOS using EMAC time series of the contribution of the different modes (Aitken mode versus accumulation mode; simulation of volcanic and dust events) as diagnostic tools.

Within the ESA Living Planet felloswhip “STatistics of AeRosol and CLouds INTeractions from satellite" (STARCLINT), a quantitative assessment of interactions between suspended aerosol particles and clouds has been carried out. Such interactions vastly determine the settings of our climate system and consequently shape human activities and well-being. The two-decadal observational record provided by satellites enables the assessment of robust statistical relationships between atmospheric particulate and clouds, despite the superimposed modulation of regional- to-continental meteorology. Long-term data sets have been used, that are generated within the ESA Aerosol, Cloud and Sea Surface Temperature (SST) Climate Change Initiative (CCI) activities, and have been sided by the output of novel developed techniques targeting the bottom altitude, the geometrical extent and the profile of droplet radius of clouds from passive hyper- and multi-spectral measurements. Together with supplementary information by complementary ground-based and space-borne records, suitable local regimes of cloud and aerosol properties are identified, the meteorology signal is singled out, and three key questions are addressed: (1) to what extent the satellite records, generated with different techniques, are representative of the properties of the same cloud parts? (2) What corrections are needed for merging time series of columnar vs vertically resolved properties? (3) what is the influence of condensation nuclei (CCN/IN) to the lifetime and the geometrical properties of cloud systems? The bottom-up observation-based approach devised within STARCLINT, together with the full length and geographical coverage of satellite records, will help in making a step toward an improved knowledge of the hydrological cycle.