From Color to Cosmos: Refining Bayesian Photometric Redshift (BPZ) for Weaklensing Mass Estimates

Understanding the structure of the universe requires accurate redshift measurements. In weaklensing mass estimates of galaxy clusters, photometric redshifts (photo-z) of galaxies are used for constructing the relative lens-source distance in the mass equation. Consequently, the accuracy of photo-z often sets the lower bound on the signal-to-noise ratio, and thus the uncertainty in mass estimation.

Photometric redshift estimation suffers from a long-standing discrepancy at low redshift due to weak color–redshift leverage and degeneracies among dust, galaxy type, and redshift. This project aims to improve photo-z calculations by incorporating the latest spectroscopic redshift (spec-z) data release from the Dark Energy Spectroscopic Instrument (DESI), extending the training set by several factors within the mass estimation pipeline. By querying and cross-matching DESI spec-z with LoVoCCS targets, we identified a systematic misclassification of low-redshift ellipticals as spirals, driven either by over-corrected extinction or by incomplete template libraries. Responding to the first hypothesis, we applied extinction-scaled flux errors and updated the pipeline’s dustmap to Planck2016, which significantly reduced the photo-z outliers.

The new algorithm reduces the photo-z outlier rate by over 50%, driving it down to ~6% for the superclusters, and has been integrated into the group’s pipeline. We will further investigate the second hypothesis by calibrating our template libraries against the The Cosmic Evolution Survey (COSMOS). Coupled with machine learning approaches, future refinements will bring us closer to a more complete and precise picture of photometric redshift and gravitational weaklensing.