Inferring the three-dimensional distribution of dust in the Galaxy with a non-parametric method: Preparing for Gaia

TL;DR

This paper introduces a non-parametric, Gaussian Process-based model to infer the 3D dust distribution in the Milky Way using stellar extinction data, enabling detailed and smooth dust maps without common artifacts.

Contribution

It develops a novel non-parametric approach employing Gaussian Processes to map 3D dust density, including unobserved regions, improving over previous methods.

Findings

Successfully reconstructs detailed dust density variations.

Produces smooth 3D dust maps without 'fingers of god' artifacts.

Validates the method with mock, simulated, and real Gaia data.

Abstract

We present a non-parametric model for inferring the three-dimensional (3D) distribution of dust density in the Milky Way. Our approach uses the extinction measured towards stars at different locations in the Galaxy at approximately known distances. Each extinction measurement is proportional to the integrated dust density along its line-of-sight. Making simple assumptions about the spatial correlation of the dust density, we can infer the most probable 3D distribution of dust across the entire observed region, including along sight lines which were not observed. This is possible because our model employs a Gaussian Process to connect all lines-of-sight. We demonstrate the capability of our model to capture detailed dust density variations using mock data as well as simulated data from the Gaia Universe Model Snapshot. We then apply our method to a sample of giant stars observed by APOGEE and Kepler to construct a 3D dust map over a small region of the Galaxy. Due to our smoothness constraint and its isotropy, we provide one of the first maps which does not show the "fingers of god" effect.