SED-dependent Galactic Extinction Prescription for Euclid and Future Cosmological Surveys

The outcome of upcoming cosmological surveys will depend on the accurate estimates of photometric redshifts. In the framework of the implementation of the photo-z algorithm for Euclid, we are exploring new avenues to improve template-fitting methods. The paper focusses on the prescription of the extinction of source light by dust in the Milky Way. Since Galactic extinction strongly correlates with wavelength and photometry is commonly obtained in broad-band filters, the amount of absorption depends on the source SED, a point often neglected as the SED is not known a-priori. A consequence of this is that the observed E(B-V) (=A_B-A_V) will be different from the E(B-V) used to normalise the absorption law k_lambda (=A_lambda/E(B-V)). Band-pass corrections are required to renormalise the law for a given SED. We assess the band-pass corrections of a range of SEDs and find they vary by up to 20%. We investigate how dust-to-reddening scaling factors depend of the sources used for their calibration. We derive scaling factors from the color excesses of z<0.4 SDSS red galaxies and show that band-pass corrections predict the observed differences. Extinction is then estimated for a range of SEDs and filters relevant to Euclid and other cosmological ground-based surveys. For high extinction line-of-sights (E(B-V)>0.1, ~8% of the Euclid survey), the variations in corrections can be ~0.1mag in the `bluer' optical filters and ~0.04mag in the NIR filters. An inaccurate correction of extinction critically affects photo-z. In particular, for high extinctions and z<0.5, the bias (mean D_z=z_phot-z_real) exceeds 0.2%(1+z), the precision required by weak-lensing analyses. Additional uncertainty on the MW extinction law further reduces the photo-z precision. We propose a new prescription of Galactic absorption for template-fitting algorithms that takes into consideration the dependence of extinction with SED.