This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| GREAT3 | |
|---|---|
| Name | GREAT3 |
| Established | 2013 |
| Type | Scientific challenge |
| Field | Astrophysics |
GREAT3 is an international blind challenge for the measurement of weak gravitational lensing shear from astronomical images. It aimed to quantify and improve methods used by collaborations preparing for surveys such as Dark Energy Survey, Euclid, Large Synoptic Survey Telescope and Wide Field Infrared Survey Telescope by providing realistic simulated images and a controlled evaluation framework. The project built on prior community efforts including Shear TEsting Programme and GREAT08, coordinating teams from observational projects, data-analysis groups, and statistical-methods researchers.
The initiative arose from concerns within Sloan Digital Sky Survey-era analyses and proposals from teams behind Hyper Suprime-Cam, Canada–France–Hawaii Telescope, Subaru Telescope, Pan-STARRS and planned missions such as Euclid and Wide Field Infrared Survey Telescope about systematic biases in shear estimation. Earlier challenges like GREAT08 and GREAT10 highlighted algorithm-dependent biases and spurred cross-field collaboration among contributors from institutions including University of California, Berkeley, University of Cambridge, Institute of Astronomy, Cambridge, Stanford University and University College London. Organizers engaged experts from Jet Propulsion Laboratory, Lawrence Berkeley National Laboratory, Max Planck Institute for Astrophysics and the European Space Agency to ensure realism in simulated instrument effects.
GREAT3 was designed to test shear-measurement algorithms against images that incorporated realistic galaxy morphology, point-spread function (PSF) variation, pixel noise, and detector effects drawn from imaging modes typical of Hubble Space Telescope, Subaru Telescope, and future facilities like Euclid and Large Synoptic Survey Telescope. The goals included benchmarking performance across methods used by teams from Kavli Institute for Cosmology, Fermi National Accelerator Laboratory, Max Planck Institute for Extraterrestrial Physics, Institut d'Astrophysique de Paris and University of Toronto; quantifying multiplicative and additive biases; and informing calibration strategies for cosmological parameter inference in projects such as Dark Energy Survey and Euclid Consortium.
Simulated data sets mimicked observations with PSF spatial variation, realistic galaxy populations from catalogues inspired by COSMOS (Cosmic Evolution Survey), and instrumental features like charge transfer inefficiency modeled after Hubble Space Telescope detectors. Images were generated using software and community codes developed at institutions such as University of Oxford, University of Edinburgh, University of Michigan, and University of Washington. The challenge included branches with "control" simulations, variable PSF branches using patterns similar to those encountered by Subaru Telescope and Canada–France–Hawaii Telescope, and space-based branches reflecting Wide Field Infrared Survey Telescope and Euclid optical designs. Training and test sets were distributed to teams affiliated with research groups at University of Pennsylvania, Columbia University, Princeton University, and Imperial College London.
Participants submitted shear estimates for multiple branches; evaluation employed metrics sensitive to multiplicative bias (m) and additive bias (c) following conventions used by collaborations including Dark Energy Survey and Kilo-Degree Survey. The scoring algorithm adapted rubric elements from GREAT08 and GREAT10 and incorporated statistical uncertainty measures used in analyses by Planck (spacecraft), Baryon Oscillation Spectroscopic Survey and Sloan Digital Sky Survey. Teams were ranked by performance across signal-to-noise regimes and PSF complexity, mirroring requirements for cosmological analyses in projects such as Euclid Consortium and Large Synoptic Survey Telescope Science Collaboration.
A wide array of methods entered, spanning model-fitting approaches developed at Max Planck Institute for Astrophysics, moment-based estimators from groups at University College London, and machine-learning frameworks from teams at Google Research, Microsoft Research, and university labs at University of Toronto and University of Cambridge. Participants included members of collaborations like Dark Energy Survey, Kilo-Degree Survey, Hyper Suprime-Cam Subaru Strategic Program, and independent groups from Stanford University, Princeton University, University of Oxford, University of Edinburgh, and Institute for Computational Cosmology. Methods incorporated pipelines inspired by codes used in Hubble Space Telescope weak-lensing analyses and techniques from statistical communities at Carnegie Mellon University and Massachusetts Institute of Technology.
Results revealed that no single method uniformly outperformed others across all branches; model-fitting algorithms tended to excel in high signal-to-noise space-based-like images, while moments-based and machine-learning techniques performed competitively in ground-based-like conditions modeled after Subaru Telescope and Canada–France–Hawaii Telescope. Systematic biases correlated with PSF complexity, galaxy morphology diversity drawn from COSMOS (Cosmic Evolution Survey) catalogues, and pixelization comparable to detectors on Hubble Space Telescope. The challenge quantified requirements for multiplicative and additive bias control necessary for missions like Euclid and Large Synoptic Survey Telescope to meet cosmological goals, echoing conclusions from teams in Dark Energy Survey and Kilo-Degree Survey.
GREAT3 accelerated improvements in shear estimation, influenced calibration pipelines in Dark Energy Survey and Euclid Consortium preparations, and guided method development for upcoming programs such as Large Synoptic Survey Telescope Science Collaboration and Wide Field Infrared Survey Telescope teams. It fostered cross-disciplinary exchanges among groups at Max Planck Society, European Space Agency, NASA, and academic institutions including University of Cambridge and Stanford University, and seeded follow-up challenges and workshops that continued to refine community standards established during earlier efforts like GREAT08 and GREAT10. The datasets and analysis lessons informed subsequent shear-calibration studies used in cosmological parameter estimation by surveys such as Dark Energy Survey and analyses by the Planck (spacecraft) collaboration.
Category:Astronomy challenges