Avoiding (photo-$z$) Catastrophe

Avoiding (photo-$z$) Catastrophe
A. J. Battisti, E. da Cunha, S. Jin, J. A. Hodge
arXiv:2507.01251v1 Announce Type: new
Abstract: Spectral modeling codes that estimate photometric redshifts (photo-$z$) are a powerful and often reliable method for determining redshifts of galaxies. However, there are notable instances where degeneracies in spectral energy distribution (SED) colors lead to `catastrophic’ failures. We highlight the case of COSBO-7, a dusty, intermediate-$z$ galaxy that masqueraded as a high-$z$ source, because it demonstrates a unique scenario where photo-$z$ codes run into issues despite extensive multi-wavelength photometry. We advocate that photo-$z$ fitting should aim to: (1) use the entire available SED (UV–radio) whenever possible to help break color degeneracies, (2) allow flexible dust attenuation prescriptions, both in terms of the attenuation curve slope and a varying 2175{AA} absorption feature, and (3) implement uncertainty floors to account for limitations in spectral models and also on the photometry itself.arXiv:2507.01251v1 Announce Type: new
Abstract: Spectral modeling codes that estimate photometric redshifts (photo-$z$) are a powerful and often reliable method for determining redshifts of galaxies. However, there are notable instances where degeneracies in spectral energy distribution (SED) colors lead to `catastrophic’ failures. We highlight the case of COSBO-7, a dusty, intermediate-$z$ galaxy that masqueraded as a high-$z$ source, because it demonstrates a unique scenario where photo-$z$ codes run into issues despite extensive multi-wavelength photometry. We advocate that photo-$z$ fitting should aim to: (1) use the entire available SED (UV–radio) whenever possible to help break color degeneracies, (2) allow flexible dust attenuation prescriptions, both in terms of the attenuation curve slope and a varying 2175{AA} absorption feature, and (3) implement uncertainty floors to account for limitations in spectral models and also on the photometry itself.