O2.3 Stacking of radio surveys

Stacking of Radio Surveys
Jeroen Stil
Department of Physics & Astronomy
University of Calgary
Stacking: piling up different sources
Statistical properties of radio emission of a sample of faint sources below the detection
limit of a survey.
Special case: Polarization of most sources not directly detectable.
Stil et al. (2014) showed sample median polarization
can be recovered to the detection limit in Stokes I.
Root-N improvement of noise up to N=105
demonstrated in NVSS.
Requires a catalog of target positions from another survey.
Special case: Stacking polarized intensity uses targets from the same radio survey.
Stack continuum or radio spectral lines
Information from stacking:
Astrophysical modeling of trends/correlations revealed by stacking related subsamples.
Example: Infrared-radio correlation for galaxies as a function of redshift.
Stacking Galaxies as a Function of Inclination
1.0 < R25 < 1.4
1.4 < R25 < 1.8
Flux Density (mJy)
17,000 targets
74 MHz
325 MHz
1.8 < R25 < 2.5
1400 MHz
Axial Ratio R25
R25 > 2.5
Stacking AGN Polarization in NVSS
Spectral Index:
Fractional polarization as a
function of 1.4 GHz flux density
Stil et al. (2014)
Cannot be done with present deep fields
Differentiate the sample by observable parameters to reveal correlations that
astrophysical models must reproduce. Beware of unintended selection
effects that may also correlate with signal strength.
Requirements for Stacking
Image cubes including “empty sky” (compromise on frequency resolution).
Uniform angular resolution and sensitivity
Flexible and efficient access to complete image archive.
Computing resources with access to data (small footprint but enduring).
Large input catalogs and advanced sample selection.
Alignment with sub-pixel accuracy for Nyquist-sampled images
Technical Challenges
Discretization of data values by design (NVSS) or by nature (X-ray photon
Tiling of survey images (overlap, gridding, sorting of images)
Copyright messages and missing data in the images
Access to metadata (survey images and target catalog)
Clustering of target sources
• Aperture-integration of intensity before stacking mitigates position errors,
resolved target sources and discretization of data
• Seeding of artificial sources to understand systematics in the data
• Stacking offset positions
• In-situ noise statistics and coordinates of peak intensity in postage stamp
• Sample statistics other than mean or median
Future of Stacking Radio Surveys
Integrate Stacking in the Archive?
Increase in resolution and bandwidth boost survey data volume (EVLA,
WSRT, ASKAP, MeerKAT, Square Kilometre Array).
Solutions to limit the cost of data storage can create significant hurdles.
Archives are not designed to retrieve millions of target sources and access
thousands of survey images simultaneously.
Use distributed science computing platforms such as CyberSKA.
Stacking radio surveys provides astrophysical information for large
samples that is otherwise inaccessible.
Broad-band surveys create new opportunities for stacking.
Current radio surveys can be downloaded and analyzed locally (NVSS,
FIRST, WENSS, ATLAS). Stacking future surveys faces challenges in data
transport and storage, unless science computing capability is integrated
with the archive.

similar documents