# Euclid preparation: X. The Euclid photometric-redshift challenge

**Authors:** N. Mauri, L. Patrizii, G. Sirri, L. Stanco, M. Tenti, Y. Wang, O. Ilbert, M. Meneghetti,  Euclid Collaboration, G. Desprez, S. Paltani, J. Coupon, I. Almosallam, A. Alvarez-Ayllon, V. Amaro, M. Brescia, M. Brodwin, S. Cavuoti, J. De Vicente-Albendea, S. Fotopoulou  
**arXiv:** [2009.12112](https://arxiv.org/abs/2009.12112)  
**Date:** 2020-09-25  
**Summary page:** https://www.summarizepaper.com/en/paper/2009.12112v1/

## Key points

- Precise and accurate photometric redshift measurements are crucial for the success of large photometric surveys for cosmology.
- No method has been able to produce photo-zs at the required accuracy using only broad-band photometry.
- The Euclid Collaboration conducted a study on the performance of 13 photometric redshift code single value redshift estimates and redshift probability distributions (PDZs) on a common set of data, focusing particularly on the 0.2--2.6 redshift range that the Euclid mission will probe.
- A challenge was designed using emulated Euclid data drawn from three photometric surveys of the COSMOS field, with a calibration sample and a validation sample containing only photometry to ensure a blinded test of methods.
- All methods could provide reliable single value estimates, but several machine-learning methods failed to produce useful PDZs.
- Participants' rejection criteria efficiently removed strong outliers where photo-z deviated by more than 0.15(1+z) from spectroscopic-redshift (spec-z).
- Further development is needed in producing accurate photo-z measurements using broad-band photometry alone for successful cosmological analyses in future large-scale surveys such as Euclid.

## Summary

The success of forthcoming large photometric surveys for cosmology depends on the ability to obtain precise and accurate photometric redshift (photo-z) measurements. However, no method has been able to produce photo-$z$s at the required accuracy using only broad-band photometry. To address this challenge, an assessment of the strengths and weaknesses of current methods is crucial. The Euclid Collaboration reports on the performance of 13 photometric redshift code single value redshift estimates and redshift probability distributions (PDZs) on a common set of data, focusing particularly on the 0.2--2.6 redshift range that the Euclid mission will probe. To conduct this study, a challenge was designed using emulated Euclid data drawn from three photometric surveys of the COSMOS field. The data were divided into two samples: one calibration sample for which photometry and redshifts were provided to participants, and a validation sample containing only photometry to ensure a blinded test of methods. Participants were invited to provide a redshift single value estimate and a PDZ for each source in the validation sample, along with a rejection flag indicating sources they considered unfit for use in cosmological analyses. The performance of each method was assessed through informative metrics using cross-matched spectroscopic and highly-accurate photometric redshifts as ground truth. The results showed that all methods could provide reliable single value estimates, but several machine-learning methods failed to produce useful PDZs. Additionally, participants' rejection criteria efficiently removed strong outliers where photo-z deviated by more than 0.15(1+z) from spectroscopic-redshift (spec-z). This study highlights the need for further development in producing accurate photo-z measurements using broad-band photometry alone for successful cosmological analyses in future large-scale surveys such as Euclid. The findings also emphasize the importance of assessing current methods' strengths and limitations towards developing approaches that meet this challenge.