theSkyNet POGS

theSkyNet POGS
NGC 2703, a nearby spiral galaxy imaged by Pan-STARRS1. Galaxies like this one were the targets that POGS volunteers helped fit pixel by pixel.
Project
Status	Completed
Category	Astronomy
Compute	CPU
Requires	None documented
Development
Developer	International Centre for Radio Astronomy Research (ICRAR), led by Associate Professor Kevin Vinsen
Author	Kevin Vinsen and David Thilker
Sponsor	ICRAR, a joint venture of Curtin University and the University of Western Australia
Maintainer	N/A (project concluded in 2018)
Initial release	September 13, 2012  (14 years ago)
Completed	May 2, 2018  (8 years ago)
Repository	https://github.com/ICRAR/boinc-magphys
Software
Written in	Python, C++, Fortran, C, PHP
Operating system	Windows, Unix/Linux, macOS, Android
BOINC statistics
Stats as of	October 13, 2014  (12 years ago)
Active users	5,268
Total users	13,770
Active hosts	16,508
Total hosts	40,847
Metadata
Website	http://pogs.theskynet.org/pogs/
License	LGPL-2.1-or-later

theSkyNet POGS (Pan-STARRS Optical Galaxy Survey) was a volunteer computing project that used BOINC to measure the physical properties of tens of thousands of nearby galaxies. It was run by the International Centre for Radio Astronomy Research (ICRAR), a joint venture of Curtin University and the University of Western Australia, as one of two research projects under the umbrella of theSkyNet, ICRAR's public astronomy outreach initiative. POGS combined imaging from three space telescopes, GALEX, Pan-STARRS1 and WISE, to build a multi-wavelength, pixel-resolved atlas of galaxy properties such as stellar mass, star formation rate and dust content. theSkyNet POGS was the first Australian-based project to be made available to the public on BOINC.

The project completed its data-processing goals in May 2018 and is no longer distributing work.

Background

theSkyNet was launched by ICRAR in September 2011 as a citizen-science initiative combining two separate research projects, Sourcefinder and POGS. Sourcefinder tested automatic radio source-finding algorithms in preparation for surveys with the Australian Square Kilometre Array Pathfinder (ASKAP) and the future Square Kilometre Array, and initially ran as a Java-based browser application before being redeveloped onto BOINC and VirtualBox. POGS, the subject of this article, began internal testing about a year after theSkyNet's launch and was formally integrated into the main theSkyNet website and opened to BOINC volunteers on 13 September 2012, coinciding with theSkyNet's second anniversary and the launch of a site-wide relaunch nicknamed "T2".

The project's name is a backronym: "POGS" originally referred to a disc-flipping playground game (see milk caps) that originated on Maui, Hawaii in the 1920s, chosen as a nod to the fact that the Pan-STARRS PS1 telescope sits on Mount Haleakala on Maui. The project subsequently recast the acronym as "Pan-STARRS Optical Galaxy Survey".

Scientific goals and methodology

POGS set out to combine the spectral coverage of GALEX (ultraviolet), Pan-STARRS1 (optical) and WISE (near-infrared) to build a value-added, multi-wavelength galaxy atlas of the nearby Universe, later also folding in data from the Sloan Digital Sky Survey (SDSS). Rather than fitting a single spectral energy distribution (SED) to the integrated light of each galaxy, the project performed SED fitting on an individual pixel basis, producing resolved maps of a galaxy's physical properties across its disc.

The twin Pan-STARRS telescopes at Haleakala Observatory, Maui, Hawaii, source of the optical imagery underpinning POGS

For each pixel, volunteers' computers fitted an observed spectral energy distribution against a large library of model galaxy spectra generated by the MAGPHYS code (Multi-wavelength Analysis of Galaxy Physical Properties), which enforces energy balance between the light absorbed by interstellar dust and the light the dust re-emits in the infrared:

$L_{dust}=\underset{0}{\overset{\mathrm{\infty }}{\int }}[L_{\lambda }^{\text{unattenuated}}-L_{\lambda }^{\text{attenuated}}]d\lambda$

The best-fitting model for each pixel was selected by minimising the chi-squared statistic between the observed and model fluxes across the available bands:

${\chi }^2=\sum _{i=1}^N\frac{{(f_i^{\text{obs}}-f_i^{\text{model}})}^2}{{\sigma }_i^2}$

From the best-fitting models, the project derived resolved maps of stellar mass surface density, star formation rate surface density, dust attenuation and first-order star formation history:

${\mathrm{\Sigma }}_{\mathrm{S}\mathrm{F}\mathrm{R}}=\frac{{\mathrm{S}\mathrm{F}\mathrm{R}}_{\mathrm{p}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l}}}{A_{\mathrm{p}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l}}}$

theSkyNet POGS relied entirely on CPU processing and did not make use of GPUs, in contrast to many contemporary BOINC astronomy projects. A GPU-accelerated version of MAGPHYS was in development as of late 2013, with testing anticipated for November of that year, though the CPU-only client remained the project's production workload for its lifetime.

Software and infrastructure

The BOINC application, published by ICRAR under the name boinc-magphys, wrapped the underlying MAGPHYS Fortran SED-fitting code inside a distributed-computing pipeline written primarily in Python and C++, with supporting components in C and PHP for the project server, database migration and web front end. The client software ran on Windows, Unix/Linux, macOS and Android, and some discrepancies were observed between results returned by Android devices and other platforms, attributed to differences in floating-point rounding at the boundary between 32-bit and 64-bit architectures. The source code is published under the GNU Lesser General Public License, version 2.1 or later.

NASA's Galaxy Evolution Explorer (GALEX) supplied the ultraviolet imagery combined into the POGS multi-wavelength atlas

The Wide-field Infrared Survey Explorer (WISE) contributed near and mid-infrared photometry to the survey

Growth and statistics

At the time of the project's founding paper (30 September 2013), POGS had 4,663 active users running 11,629 computers, and had processed 8,348 galaxies, equivalent to about 53 million individual pixels, at a pace of roughly 110 galaxies per day. On 23 September 2014 the project's team leader announced that POGS was about to process its 50,000th galaxy. The project's server status page recorded, as of 13 October 2014, a total of 13,770 users who had ever earned credit (5,268 of them recently active) running a total of 40,847 computers with credit (16,508 recently active).

Completion and legacy

By late April 2018, volunteers noticed that no new POGS work units were being issued. Team leadership confirmed shortly afterwards that all of the roughly 100,000 target galaxies identified for the survey had been processed and that the project was concluding. The team indicated that the processed pixel data would next be used to train an autoencoder neural network to help classify galaxies by their measured features, and that any resulting scientific papers would credit the BOINC volunteers who contributed processing time. Keeping the theSkyNet and POGS web services online was reported to cost more than 2,000 AUD per month, and the team indicated the sites would eventually be taken offline once the wrap-up work was finished. ICRAR's sister project theSkyNet Sourcefinder was shut down for redevelopment around the same period. The two projects together are now listed among the completed and retired BOINC projects.

Following the closure of theSkyNet, ICRAR went on to run further citizen-science projects, including AstroQuest, launched in 2019 to help classify light sources within galaxy images.

Publications

The following papers, meeting the BOINC publications criterion of containing scientific results arising, directly or indirectly, from BOINC-based computing, are associated with theSkyNet POGS.

1. (2016).The Pan-STARRS1 Surveys. DOI: 10.48550/ARXIV.1612.05560.
2. (2014).Early science from the Pan-STARRS1 Optical Galaxy Survey (POGS): Maps of stellar mass and star formation rate surface density obtained from distributed-computing pixel-SED fitting.
3. (2013).A BOINC based, citizen-science project for pixel spectral energy distribution fitting of resolved galaxies in multi-wavelength surveys. Astronomy and Computing. pp. 1Template:Ndash12. DOI: 10.1016/j.ascom.2013.10.001.

See also

• SETI@home
• Einstein@Home
• World Community Grid
• Predictor@home