GALEX-SDSS photometric redshifts

Sébastien Heinis, Stéphane Arnouts, Tamás Budavári

sebastien@pha.jhu.edu

The sample

We use here a subsample of MIS fields from GR2 with NUV exposure times greater than 1000s and SDSS counterparts (GR5) within 0.5deg from the center (348 fields in total); we refer to this dataset hereafter, unless otherwise stated.



In the following, figures use:

• Fields with exposition times in NUV or FUV greater than 1000 s.
• Objects within the GALEX primary resolution
• Non saturated objects in SDSS bands
• Dereddened magnitudes according to Schlegel et al maps using:
  dered_fuv = fuv_mag - 8.24*e_bv
  dered_nuv = nuv_mag - (8.20*e_bv  -0.62*e_bv*e_bv)
  
• NUV or FUV magnitudes brighter than 22
• Objects non masked in SDSS (BLEEDING, BRIGHT_STAR, TRAIL and HOLE) or GALEX (flag > 1, and nuv_mag < 16.5) masks
• The following color-code: blue for galaxies, red for stars, and green for qsos

The Method

• A template-fitting based method, using S. Arnouts code, called LePhare . We used here the following templates:
  • Galaxies: 42 SEDs from Bruzual & Charlot (2003) (42 GISSEL)
  • Stars: SEDs from Pickles (1998), Michel Laget (p. comm.) and 4 White Dwarves spectra (Bohlin et al, 1995)
  • QSOs: Synthetic spectra from Cristiani & Vio (1990) and Cristiani et al (2004)
• An empirical magnitude-redshift relation, based on Connolly et al (1995), using an implementation from I. Csabai (called Polyfit)

1. Calibration of the redshift-magnitude relation

   The first step consists to calibrate a redshift-apparent magnitude relation based on the GALEX NUV band and the SDSS spectroscopic counterparts. We use the 6 bands (NUV, u, g, r, i, z) and the spectroscopic redshift of SDSS galaxies to fit a 3rd degree polynom: z = f(NUV, u, g, r, i, z).
   We use for the training only GALEX sources with a spectroscopic SDSS counterparts and a 1 to 1 GALEX-SDSS cross match within a 4 arcsec radius. Results are not insensitive to this selection (see Training set).
   The polynom coefficients are hereafter used to derive a photometric redshift. Note that this calibration rely on the SDSS spectroscopic sample (selected with r < 17.5), and that the relation derived is, strictly speaking, only valid in the same volume. This means that it is safest to use these photometric resdhifts with z ~ < 0.25. The following figures explicitly use them in a larger volume.

2. Template-fitting with no assumption

   We first compute photometric redshifts using LePhare with redshift as a free parameter. The code use galaxy, star, and qso templates.

3. Template-fitting with Polyfit redshift assumption

   We then compute photometric redshift using LePhare fixing redshift as the one derived from Polyfit. The code only uses galaxy templates.

4. Template fitting with spectroscopic redshift assumption

   We finally compute photometric redshift using LePhare fixing redshift as the spectroscopic one. The code only uses galaxy templates.

The results

Table description