Survey of Surveys

II. Stellar parameters for 23 million stars

¹ INAF – Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, 50125 Firenze, Italy
² Space Science Data Center, Via del Politecnico SNC, 00133 Roma, Italy
³ INAF – Osservatorio Astronomico di Roma, Via Frascati 33, 00040, Monte Porzio Catone, Roma, Italy
⁴ Institute of Astronomy, 48 Pyatnitskaya St., 119017 Moscow, Russia
⁵ Dipartimento di Fisica e Astronomia, Università di Firenze, Via G. Sansone 1, 50019 Sesto Fiorentino, FI, Italy
⁶ Instituto de Astrofísica de Canarias, 38205 La Laguna, Tenerife, Spain
⁷ Universidad de La Laguna, Dpto. Astrofísica, 38206 La Laguna, Tenerife, Spain
⁸ Université Côte d’Azur, Observatoire de la Côte d’Azur, CNRS, Laboratoire Lagrange, Nice, France
⁹ Zentrum für Astronomie der Universität Heidelberg, Landessternwarte, Königstuhl 12, 69117 Heidelberg, Germany
¹⁰ Max Planck Institute for Astronomy, Königstuhl 17, 69117 Heidelberg, Germany
¹¹ Leibniz-Institut für Astrophysik Potsdam (AIP), An der Sternwarte 16, 14482 Potsdam, Germany

^★ Corresponding author.

Received: 28 May 2025
Accepted: 2 July 2025

Abstract

Context. In the current panorama of large surveys, the vast amount of data that are obtained with different methods, data types, formats, and stellar samples prevents an efficient use of the available information.

Aims. The Survey of Surveys is a project to critically compile survey results into a single catalog to facilitate the scientific use of the available information. In this second release, we present two new catalogs of stellar parameters (T_eff, log g, and [Fe/H]).

Methods. To build the first catalog, SoS-Spectro, we internally and externally calibrated stellar parameters from five spectroscopic surveys (APOGEE, GALAH, Gaia-ESO, RAVE, and LAMOST). Our external calibration on the PASTEL database of high-resolution spectroscopy ensures better performances for data of metal-poor red giants. The second catalog, SoS-ML catalog, is obtained by using SoS-Spectro as a reference to train a multilayer perceptron that predicts stellar parameters based on two photometric surveys, SDSS and SkyMapper. As a novel approach, we built on previous parameter sets from Gaia DR3 and other sources to improve their precision and accuracy.

Results. We obtained a catalog of stellar parameters for about 23 million stars that we make publicly available. We validated our results with several comparisons with other machine-learning catalogs, stellar clusters, and astroseismic samples. We found substantial improvements in the parameter estimates compared to other machine-learning methods in terms of precision and accuracy, especially in the metal-poor range. This was particularly evident when we validated our results with globular clusters.

Conclusions. Our results at the low-metallicity end improve for two reasons: First, we used a reference catalog (the SoS-Spectro) that was calibrated using high-resolution spectroscopic data; and second, we chose to build on pre-existing parameter estimates from Gaia and Andrae et al. and did not attempt to obtain our predictions from survey data alone.