Multivariate Time-series Analysis of Variable Objects in the Gaia Mission

TL;DR

This paper explores the analysis of multivariate time-series data from the Gaia mission, focusing on extracting astrophysical information through advanced statistical techniques to improve data characterization.

Contribution

It introduces methods for analyzing Gaia's diverse multi-epoch measurements and demonstrates how these techniques enhance astrophysical data extraction.

Findings

Assessment of Gaia's time sampling and measurement types

Application of principal component analysis to Gaia data

Development of multi-response regression methods

Abstract

In astronomy, we are witnessing an enormous increase in the number of source detections, precision, and diversity of measurements. Additionally, multi-epoch data is becoming the norm, making time-series analyses an important aspect of current astronomy. The Gaia mission is an outstanding example of a multi-epoch survey that provides measurements in a large diversity of domains, with its broad-band photometry; spectrophotometry in blue and red (used to derive astrophysical parameters); spectroscopy (employed to infer radial velocities, v sin(i), and other astrophysical parameters); and its extremely precise astrometry. Most of all that information is provided for sources covering the entire sky. Here, we present several properties related to the Gaia time series, such as the time sampling; the different types of measurements; the Gaia G, G BP and G RP-band photometry; and Gaia-inspired studies using the CORrelation-RAdial-VELocities data to assess the potential of the information on the radial velocity, the FWHM, and the contrast of the cross-correlation function. We also present techniques (which are used or are under development) that optimize the extraction of astrophysical information from the different instruments of Gaia, such as the principal component analysis and the multi-response regression. The detailed understanding of the behavior of the observed phenomena in the various measurement domains can lead to richer and more precise characterization of the Gaia data, including the definition of more informative attributes that serve as input to (our) machine-learning algorithms.