Skip to main content

Development of Novel Statistical Tools for the Analysis of Astronomical Data

Periodic Reporting for period 2 - ASTROSTAT (Development of Novel Statistical Tools for the Analysis of Astronomical Data)

Reporting period: 2018-01-01 to 2019-12-31

Astrophysics is an ancient science which in the past 50 years has seen an explosion in the volume of available data. This is thanks to the construction of telescopes and the launch of satellites with state of the art instruments that allow the study of extremely faint objects from the Milky Way to in the very early Universe. In parallel Statistics, a major branch of Mathematics - another ancient science - has seen major advances in the past century with the development of novel methods that have revolutionized the analysis of large volumes of data. This project combines the power of modern statistical methods with state of the art astronomical data and simulations, in order to solve outstanding problems in Astrophysics through the development of novel statistical tools. Our efforts focus on problems related to:
(i) The classification of objects detected in large and/or deep surveys of the sky with ground and space-based observatories
(ii) The characterization of the populations of stars and galaxies observed in these surveys
(iii) The characterization of faint and/or blended objects in optical and X-ray observations
(iv) Modelling the cosmological evolution of the supernovae that are used as probes of the expansion of the Universe
(v) Analysis of astronomical data with complex multi-dimensional models

Addressing these problems is key for solving long-standing questions that are in the forefront of modern Astrophysics, including the cosmological evolution of galaxies and their stellar populations, the formation and evolution of binary stellar systems containing black holes or neutron stars, and the expansion of the Universe. The complexity of these problems and the large volume of diverse, multi-observatory data required entail the development of advanced analytic methods.

In order to achieve these goals we have formed a broad international network between 3 European institutions (Foundation for Research and Technology Hellas; Imperial College; Geneva Observatory) and 4 US-based institutes (Smithsonian Astrophysical Observatory; Northwestern University; University of California, Davis; Ohio State University). This network involves leading groups in Astrophysics and Statistics. The statistical methods developed as part of this effort will be used to address long-standing questions, while at the same time training the next generation of researchers in state-of-the-art statistical techniques. Such techniques are now commonly used in a wide variety of applications spanning virtually all areas of science: e.g. biology, discovery of new drugs and treatments, economics, weather forecasting, disaster management, and computer science (internet analytics). Therefore, this program has far reaching implications, either directly (through the development of tools and methods that are readily used in other disciplines) or indirectly by training young researchers to become fluent in modern data management and analysis techniques.
The premise of this program is to develop novel statistical methods to address key questions in astrophysics by enabling the close collaboration of Astrophysicists and statisticians in leading research institutes in Europe and the US. During the course of this program we embarked in a diverse range of projects covering areas such as:
- Classification of galaxies on the basis of their activity.
- Classification of stars according to their spectra types.
- Classification of young stellar objects.
- Characterization of the stellar remnants in X-ray binary stellar systems observed in nearby galaxies.
- Determination of the significance of features observed in one and two dimensional data.
- Identification of structures in (noisy) images.
- Analysis of flux distributions of astronomical sources.
- Formation and evolution of accreting binary stellar systems.
- Formation and evolution of binary black-hole systems.
- Direct and indirect detection of dark matter particles.
- Use of type Ia Supernovae as cosmological probes.

In addition to these planned projects we also initiated new ones that resulted from the collaboration and exchange of ideas between the consortium partners. These include:
- Study of the age of stellar populations.
- Determination of scaling relations between galactic X-ray emission and galaxy stellar population parameters.
- Determination of the differential emission measure of stars.
- Analysis of spectra from superheated plasmas.
- Development of a method for the characterization of supernova remnants.

For these projects we have either published papers presenting the methods we have developed and the corresponding results (including relevant software tools), or we have completed the development of the methods and we are in the process of producing the relevant publications.
All methods developed as part of this program address cutting edge problems that require the development of new analysis and inference tools. By the end of the program we have developed a diverse portfolio of data analysis techniques. These techniques address three main categories of astrophysical problems:
-- Classification of astronomical objects.
-- Analysis of images and feature detection.
-- Fitting complex data with multi-dimensional, multi-layered models. These include modeling binary stellar systems and their end-points, derivation of cosmological parameters, studies of the cosmological evolution of galaxies and black holes.
The tools developed to address these problems are available to the astronomical community by the end of the program.

The most important impact of this program is the transfer of knowledge between research groups active in very different areas of science, and the training of Astrophysicists and Statisticians to use novel, advanced statistical analysis and inference methods. This will pave the way for new discoveries and the the efficient exploitation of the large volume of astronomical data that will be produced by future surveys. In addition several of these methods are relevant to areas beyond Astrophysics. Furthermore, the training of researchers in advanced data analysis techniques gives them a diverse set of skills as well as the opportunity to follow new career paths.
Study of a solar active region (Yu et al. 2018 ApJ 866 146; doi:10.3847/1538-4357/aadfdd)
Detection of jet structures in X-ray (Chandra) images (from McKeough et al. 2016; 10.3847/1538-4357)
Detection of shell-like features on galaxies using machine-learning methods(Bonfini et al, in prep.)
Evolution of galaxies on the fine-structure mass-deficit plane (Bonfini et al; 10.1093/mnras/slx169)
Detection of extended regions on X-ray (Chandra) data using seeded region growing (Fan etal in prep)