How is interstellar gas density mapped from GAIA data?

Question

I found the image below in Space.com's article This 3D Color Map of 1.7 Billion Stars in the Milky Way Is the Best Ever Made.

The caption for this image reads:

The Gaia spacecraft gathered observations for this all-sky view of the Milky Way and neighboring galaxies between July 2014 and May 2016, releasing the data on April 25, 2018. This image shows all the stars' colors and brightness (top), the total density of stars (middle) and the distribution of interstellar gas and dust across the galaxy (bottom). Credit: ESA/Gaia/DPAC

The first two maps reflect data obtained from analysis of GAIA's measurement of starlight as explained in this answer, but the third map is described as "the distribution of interstellar gas and dust across the galaxy."

As far as I understand, GAIA does not transmit all complete image data to Earth, but instead does a significant amount of image processing, object detection, classification and analysis, and other data reduction in order to extract positions and radial velocity measurements for point-like objects such as stars, though of course small solar system bodies such as comets and asteroids end up in the raw data as well.

Question: How is interstellar gas density obtained from GAIA data?

Answer 1

Two techniques immediately spring to mind.

For the stars you detect, you can compare their colours and luminosities (Gaia provides photometric colours and distances) with what you expect for a star of that type at that distance. The difference between what you expect and what you observe tells you the reddening and extinction caused by interstellar dust, integrated in a column towards that star. By going through this process for lots of stars in different directions you build up a 3D picture of the distribution of dust.

The measurements are mostly insensitive to gas - simple extrapolation between dust and gas is often assumed.

A second technique would be just to compare how many stars you can see and with what brightness distribution, with a model for the Galactic stellar population and dust distribution. This is an old-fashioned way of estimating dust extinction, works well locally, but I suspect not what is used in that map, for the simple reasons that Gaia also provides the distances to stars, and one of the purposes of the mission is to improve the Galactic model, not assume it.