Aerosol for Copernicus EUMETSAT Swansea (ACES)

This Copernicus study is funded by the European Commission (EC), procured under EUMETSAT leadership, and led by the Aerosol team of Prof. Dr Peter North from Swansea University (SU). It aims to support further understanding of, and potential improvements to, the algorithm implemented in the preliminary operational processor for the Copernicus Sentinel-3 (S3) Near Real Time (NRT < 3h) Aerosol Optical Depth (AOD) product from the Sea and Land Surface Temperature Radiometer (SLSTR). The current algorithm version is a joint development between SU and EUMETSAT. Part of the legacy developments also come from past dual-view aerosol developments funded and procured by the European Space Agency (ESA) aerosol Climate Change initiative (CCI).

The overall goal is to investigate three new questions emerging from feedback from EUMETSAT and the operational user community: e.g. the Copernicus Atmospheric Monitoring Service (CAMS). These investigations are directly focused on the latest version of the independent SU algorithm prototype.

Atmospheric aerosols originate from sources such as wildfires, emissions from pollution, sea salt and desert dust. They have a net cooling effect on climate, but with high uncertainty, and are also a major factor in air quality. Satellite retrieval is needed to improve understanding of the role of aerosols in climate, and for assimilation in models of atmospheric transport.

While both surface reflectance and atmospheric aerosol are needed independently from satellite observations of reflected sunlight, any such measurements include a mixture of contributions from both aerosol and albedo. Research by Swansea University developed new methods that accurately separate atmospheric from surface signals in satellite images using physical model inversion techniques. In collaboration with additional EUMETSAT developments, this has been developed into an NRT aerosol product from the SLSTR instruments on the Copernicus Sentinel-3 satellite series, with two satellites giving near daily global coverage. This project aims to improve aspects of this algorithm, and further develop fundamental understanding of the aerosol retrieval.

Objectives

There are three objectives to the planned research:

1. AOD retrieval over water surfaces:
   The goal of this task is to investigate the quality of the AOD retrieval from the SU prototype over sea and ocean surfaces. The emphasis is on further understanding the exact reasons why SLSTR AODs are currently lower than AOD from the MODIS/VIIRS/PMAP satellite products over global oceans, as demonstrated by EUMETSAT and CAMS results.
2. Land surface reflectance model:
   The retrieval over land incudes inversion of a general model for land surface bidirectional reflectance distribution function (BRDF). The goal of this task is to re-evaluate the accuracy of the land surface BRDF model for clear-sky cases considering all possible geometries encountered today with the SLSTR dual-view configuration.
3. Optimisation scheme:
   The goal of this task is to thoroughly review the overall inverse modelling scheme and associated cost function penalties, as implemented in the SU prototype.

Overview

This scientific study aims to support the Copernicus Sentinel-3 (S3) Near Real Time (NRT) Aerosol Optical Depth (AOD) product from the Sea and Land Surface Temperature Radiometer (SLSTR). The context is the on-going developments by EUMETSAT to continue the evolution of the dedicated Copernicus S3 NRT AOD processor. The main goal of this activity is to investigate a series of scientific elements, structured around 3 key tasks based on the existing Swansea University SLSTR aerosol prototype. Any elements demonstrated to be successful from the delivered SU SLSTR aerosol prototype could then be transferred to the processor directly by EUMETSAT in due time.

As part of the Copernicus programme, funded by the European Commission, Sentinel-3 is the third of the Sentinel satellite series, originally dedicated to land and ocean applications including sea-ice, water pollution monitoring in open-ocean and coastal areas, surface temperature, sea height, and vegetation productivity. The first satellite, Sentinel-3 A, has been flying since 16 February 2016. The second satellite, Sentinel-3 B, was successfully launched on 25 April 2018.

The potential for an aerosol product was demonstrated during the precursor missions, the (A)TSR instruments on ERS-2 and ENVISAT, and Swansea University developed a successful product under the ESA Climate Change Initiative for (A)ATSR, and developed this further as a product for SLSTR on Sentinel-3. EUMETSAT is responsible for undertaking the processing and dissemination of the S3 NRT AOD product, with a first version (v1.0) based on this algorithm. While giving good results over ocean, initial results over land showed less accuracy than the (A)ATSR product. Both EUMETSAT and SU have subsequently and independently developed this algorithm further. Figure 1 shows example AOD retrieved using the current SU prototype for March 2019 from S3A SLSTR.

One of the major geometry configuration differences between the (A)TSR instruments and SLSTR is the viewing direction of the oblique view: while the (A)TSR oblique view was pointing forward w.r.t. the satellite motion, the SLSTR oblique view is backwards w.r.t the satellite motion (respectively South and North on the descending part of the satellite orbit). While such a rotation has a priori no impacts on thermal infrared measurements and derived L2 products, such as Sea Surface Temperature (SST), it directly affects the spatial distribution of the viewing configuration associated with radiance measurements acquired with the solar channels. Additional significant differences include large extension of the swath width for both views, and two additional spectral channels in the Short Wave InfraRed (SWIR) at 1.375 (cf. Cirrus band) and 2.25 µm.

Key questions to be addressed by this research are:

1. The AOD over ocean is normally retrieved with higher accuracy than land due to the more predictable surface signal. However, the retrievals from NRT SLSTR show lower values than some comparable products (MODIS, VIIRS, PMAP). The project will investigate reasons for differences, and compare with accurate ground-based measurements from AERONET and MAN.
2. Over land the SU prototype utilises a parameterised model of surface BRDF, allowing retrieval over all surfaces free of snow, ice and cloud without further assumptions on spectral properties using the dual-view capability of the instrument. However, the constraint gives accuracy which varies with solar/view geometry, and generally results in more accurate retrieval over Southern hemisphere land than Northern for SLSTR. This research will examine this accuracy, and explore further constraints to improve AOD retrieval especially in current difficult cases over bright desert and Northern latitudes.
3. The method uses a constrained numerical inversion of a cost function to retrieve atmospheric and surface parameters, with propagation of error from observation and model uncertainties to give a per retrieval uncertainty. This research will re-evaluate and improve the inversion scheme if possible, especially considering stability and convergence, and use of most recent channel radiance uncertainties.

The project will focus on the latest prototype version developed by Swansea University, and will make this code and revised Algorithm Theoretical Basis Document available to EUMETSAT to support ideas for improvements in forthcoming NRT product development activities.