Pillar E: Astronomy and space-situational awareness
Spokesperson: Marco de Vos (ASTRON)
Deputy: Albert-Jan Boonstra (ASTRON)
Driven by scientific curiosity and technological advances, astronomical telescopes create ever increasing amounts of data. In the radio domain for example, the LOFAR ILT radio telescope array (see picture for participating countries) produces roughly 10 terabyte daily, of which 7 peta byte of condensed data is stored annually. The first phase of the SKA radio telescope, to be constructed in the coming decade, will even produce about 300 petabyte annually. Science archives can be big multipliers for total science output, provided there is easy access to these datasets and provided adequate processing capabilities are available. This is true not only for the science that the data were originally produced for, but also for related and more distant scientific fields.
In astronomy and space sciences, as in most other scientific fields, people are working towards providing easy and FAIR access to data, and open access to publicly funded research results. One golden rule in FAIR data principles (cf. Go Fair, www.go-fair.org) is to find and define the smallest common denominator that can describe the data. This core data description should be complex enough to serve several goals in one or more scientific fields, but simple enough for users to be able to connect to.
This FAIR Data Infrastructure pillar D aims to support space situational awareness sciences, solar physics, and planetary sciences. These fields are combined as they have big overlaps in terms of physical phenomena, and observational and calibration strategies. The figure below shows an example of such data: solar dynamic spectra in the form of a spectrogram. It shows solar radio emissions as received with LOFAR.
Research and development in pillar D will be done in close collaboration with the scientific fields. For the complexity reason mentioned above, this pillar will set out with a combined FAIR focus for these selected scientific fields, but will remain connected with the broader (radio) astronomy community.
Within fairdi.eu, Pillar D will address the following topics:
- Streaming data architectures, accelerators, and IoT. This topic addresses optimal data processing and storage architectures in terms of resource usage and response time, given the vast amounts of (streaming) data that needs to be processed for LOFAR space weather applications.
- Streaming Data Analytics and Compression. This topic addresses generic tools for streaming data analytics and data processing that can automatically find anomalies and cleverly compress the data on the fly, preventing the loss of relevant information.
- Generative Adversarial Neural Networks for Producing Realistic Samples. This latest major breakthrough in artificial intelligence provides a unique generic framework for training models to produce much more realistic-looking data.
- Learning from Users of Virtual Research Environments. Logs of users’ search behaviour and their analysis steps will provide a wealth of information that can be mined for better accessibility and effectiveness of the database as a whole.
- Integrating Machine Learning and Predictive Simulation. Machine learning can be used to find patterns (mapping) between simulation’s input parameters and the generated outputs. In radio astronomy these techniques in principle can be used to efficiently calibrate or to ‘switch’ between the ‘aperture plane data’ and the ‘image domain data’.
- Connection to user needs. There are many (potential) user communities that could benefit from LOFAR Space Situational Awareness R&D, and which need to be consulted.
- Last but not least: FAIR. This topic addresses the FAIR aspects of the pillar.
The topics mentioned above are described in terms of their relevance for pillar D. However, the underlying techniques for a majority of the mentioned topics have a big overlap with the goals and approaches of the other pillars.