Team: Ozone and Air Quality
Most spectrometers for satellite-supported remote sensing of the atmosphere are passive instruments. They register the sunlight reflected back and scattered into space by Earth's atmosphere. Spectral differences compared with the incoming irradiance allow conclusions about the state of the atmosphere, particularly its chemical composition.
The group “Ozone and Air Quality” of DLR's EOC has worked with UV-Visible-Near-Infrared (UVN) sensors and the associated retrieval methodologies since GOME (1995-2011) on ERS-2, Europe's first atmosphere spectrometer in orbit. Higher performance instruments followed with SCIAMACHY (2002-2012) on ENVISAT and the presently operating GOME-2A (2006-2020), GOME-2B (since 2012) and GOME-2C (since 2020) on MetOp platforms. The era of Copernicus atmosphere missions began in October 2017 with the launch of TROPOMI on the Sentinel-5 Precursor Mission. The future Copernicus missions Sentinel-4 (mid 2025) and Sentinel-5 (end 2025) will extend these global atmosphere measurements for the next two decades. The Copernicus missions are characterised by very high spatial resolution w.r.t. their predecessor missions. This leads to a hundredfold increase in the amount of data to be managed and poses great challenges for the retrieval algorithms used and the development of the resulting operational software systems.
These software systems, called processors, have the task of rapidly and continuously generating precisely-defined satellite products from measurement data. They function in an operational 24/7 environment without interactive intervention, which puts high demands on the developed software. Such software is always the end result of a long development chain. At the beginning there are retrieval algorithms that must be adapted to meet the requirements of the ground segment. In the case of Level 2 processors the starting point is the scientific retrieval code. Here, the often complex algorithm has to be so optimised that it not only provides results with the specified precision, but also itself requires as few resources as possible and runs in a stable and robust way.
Operational processors are required to generate data in Near Real-Time (NRT, usually the Level 2 data products are supposed to be ready three hours after sensing of the raw satellite data) or, in an offline mode, with the highest possible precision, without the strict timeliness requirements as in the NRT case. Sometimes the algorithmic essence of both kinds of processors is basically identical. In other cases, because of the strict time constraints of NRT operation, specially customized retrieval algorithms are required. In addition to the specifications for operational processors, a science prototype is created for every processing level for purposes of development and testing. It allows interactive control and, as required, detailed investigation of each individual processing step. Which algorithms are incorporated in the processors is described in an Algorithm and Technical Baseline Document (ATBD). These ATBDs, together with other specifications, forms the foundation for the mission operator to implement the processor software which will be finally installed at the ground segment.
Processors and their underlying retrieval algorithms further evolve in the course of a mission. The mission-specific products are regularly validated with ground measurements as a reference. As the length of a mission increases, instrument calibration improves as does the ability to extract from the measurement data already defined or even new geophysical parameters. For that reason, ideally all the measurement data is reprocessed at regular intervals using the current state-of-the-art algorithms implemented in the operational software. To this extent, the responsibility given to IMF for retrieval algorithms and processors for UVN spectrometers means long-term involvement that accompanies a mission from its conception to the time after its operation in orbit.
For the UVN sensors coming from the GOME series, we developed with UPAS, (Universal Processor for UVN Atmospheric Spectrometers) a generic processor that makes it possible to retrieve atmosphere parameters independent of sensor type. With UMAS, the Universal Mapping tool for Atmospheric Spectrometers, we provide the retrieved atmospheric parameters on daily global maps on a regular grid of the Earth’s surface as so-called Level 3 data. In order to successfully meet the high requirements of the new Copernicus atmosphere missions, state-of-the-art technologies like artificial intelligence (AI) and machine learning (ML) are being applied. .
The atmospheric trace gases products (ozone, nitrogen dioxide, sulphur dioxide, formaldehyde, bromine oxides, glyoxal, water vapour, etc.) generated with UPAS, as well as aerosol and cloud properties are being used in numerous applications including climate research, air quality and volcano monitoring and are being assimilated in various forecasting and monitoring systems (e.g. CAMS). Especially our continuous and global monitoring of the aforementioned atmospheric species as indicators for air quality is essential to characterize local as well as global trends, to understand the underlying physical processes better and to provide a solid scientific fundament to assess potential mitigation approaches to reduce air pollution. In this context, our data flow into so-called essential climate variables (ECVs), e.g. long-term data records for ozone, which are also being used by international organizations like the World Meteorological Organization (WMO).
The IMF team “Ozone and Air Quality” plays a leading role worldwide in developing retrieval algorithms and processors for the current atmosphere missions GOME-2/MetOp and TROPOMI/Sentinel-5 Precursor as well as for the future Sentinel-4 and Sentinel-5 Copernicus missions.