Type: Talk
Session: Photometric Redshifts
Author: Noah Weaverdyck

Abstract: Existing photo-z estimators are subject to a wide range of systematic effects, from modeling choices to the base "truth" sets used to train the estimator. We propose a direct calibration approach (DirCal) that is not subject to most of these systematics as a complementary means to estimate the redshift distribution of a photometric sample. As proof-of-concept, we estimate the n(z) of 3 of 4 lens bins in the Dark Energy Survey (DES) using a single bright-time pointing with $\sim3000$ spectra from the Dark Energy Spectroscopic Instrument (DESI), and compare to published results. Even further gains can be made using the information gleaned from direct calibration on "unknown unknowns" to prune the photometric sample, removing galaxies that are most likely to be problematic with minimal impact on statistical power. Such an approach promises to be a useful tool for leveraging the full statistical power of Rubin, which lies largely outside the support of existing spectroscopic datasets.