PSF star selection notes

[sfarrens]

The objective of this issue is to gather notes on the various methods for selecting stars to modelling a PSF discussed during the reading group.

[sfarrens]

@andrevitorelli's notes in Google doc

[sfarrens]

@sfarrens's notes:

Star Selection

• CFIS (PSFEx & MCCD)
  • Cut in magnitude-size (FWHM) plane
  • Assume size does not depend on luminosity (stellar locus)
  • Set range around the mode of the distribution to select stars
  • Difficult to distinguish faint galaxies from stars
  • Make a magnitude cut to remove these objects
  • This is the input of the PSF model
  • Some bright stars appear larger, cut at low mag to avoid this
  • brighter-fatter?
    • not a problem for MegaCAM
  • MCCD rejects outliers in terms of shape or size
• LensFIT
  • Strongly dependent in star rejection
  • Start with a large sample of stars
• HSC
  • Modified PSFEx (no star selection)
  • Need to address brighter-fatter effect
  • K-means selection of stellar locus
  • Selection done per CCD (~72 stars/CCD)
  • Method sensitive to seeing
• KiDS
  • Friends-of-friends
  • 2nd and 4th order moments of all objects
  • 5 dithers, hence 5 points of measure for each star
  • Combine all exposures
  • Needs to be selected as a star in at least 3 exposures
  • Performs better than CFHTLenS, which used co-added images
  • claim <1% outliers
  • empirically determined linking parameters
  • skewness and kurtosis of light profile, stars should be peaky
• DES
  • Size-magnitude plane
  • Pipeline includes a brighter-fatter effect correction
  • Flat stellar locus due to correction
  • Can include lower magnitude stars
  • Good match to Gaia
• Other methods
  • Size-magnitude + colour information
• Brighter-Fatter
  • Not all cameras suffer from this effect
  • Thicker the CCD, the stronger the effect
  • Affects all stars, strongest for brightest stars
  • Correction is robust, but takes a lot of work to implement
• Completeness
  • Some surveys require more than others to model the PSF
  • Euclid wants 100 stars/CCD
  • Purity more important
  • Gaia has enough star coverage for many surveys
    • Perhaps not deep enough
    • Also not perfect
    • Parallax information
    • Very small astrometric error
    • Match stars to surveys or simply take Gaia positions and create postage stamps
  • Remove CCDs with fewer stars than the requirement to model the PSF
  • Zhang et al. (2019)
  • Ma et al. (2008)

[sfarrens]

Discussion, how to structure star selection in the book.

We agreed that the focus should be on the methods rather than surveys. We can cite papers and mention surveys with regards to specific details in each section.

Proposed Subsections

1. Subsection on size-magnitude plane and how to identify the stellar locus
   • This section will be longer than the others and include more detail
   • Important to include examples of both clear and messy loci
2. Fourth order moments-based methods
3. Colour-based methods
4. Experimental methods (ML from pixels)
   • Also the use of GAIA

Important Details

• Emphasise the importance of sample purity and this is why simple but trusted methods are often employed

Minor Details

• Binary stars