Negative values in radio flux maps? I am trying to understand some maps of radio emission in e.g the Milky Way. I continue to see in papers that these radio maps report negative values of the flux in Jansky units, although no explanation is given as to why.
For example Figure 1 here: https://arxiv.org/pdf/2201.10541.pdf and here: https://arxiv.org/pdf/2107.02695.pdf

I am wondering if the negative values are simply due to background subtraction, or if there is something more complicated going on.
If it's just background subtraction, a simple reference explainaing how the background subtraction is done in general would be useful.
 A: 
I am wondering if the negative values are simply due to background subtraction, or if there is something more complicated going on.

In fact, there are a lot of complicated things going on here.
You have provided links to two papers that rely on interferometric observations from the Very Large Array (VLA), and MeerKAT.
VLA and MeerKAT are astronomical interferometers, i.e. telescope arrays acting together as one single virtual telescope. This implies a specific process of imaging, which is (mainly) responsible for the negative values in your maps.
This gif shows how sky images are produced by arrays of telescopes thanks to aperture synthesis:

During the first part of the gif, we are only adding more telescopes to the uv plane (which represents the ground on which telescopes are positioned), then the rotation of the sky takes care of the rest.
Since the sky image is sampled by a fringe pattern, we are bound to  have artifacts and negative values in the final map, although this is minimized by adding more telescopes to the uv plane. And there are also algorithms that are supposed to "clean" (reconstruct) the map.
In the papers that you linked, the authors quickly mention the process of cleaning:

Natural weighting was employed in the CASA task tclean to recover the best signal/noise ratio possible at the expense of resolution.


All imaging was done using wsclean (Offringa et al. 2014) with multiscale cleaning (Offringa & Smirnov 2017) enabled, and a Briggs (1995) robust value of −1.5, to provide high (∼4′′) angular resolution and suppress the sidelobes of the synthesised beam.

There are many techniques used to clean an interferometric map, including iterative algorithms, and/or using a source model.
Depending on the signal-to-noise ratio of the observations, on the size and configuration of the array, and  on the quality of the cleaning procedure, an interferometric maps can contain a lot of negative values, such as in your example.
Lastly, an other important parameter is the use of short-spacings observations. When using an interferometer (such as ALMA), we are mainly sampling high-resolution features, since the interferometer acts as a filter that ignores small spatial frequencies.
That is why it is useful to combine an interferometric map with a single-dish map, obtained with a regular radio telescope, such as the IRAM-30m.
Using short-spacings observations would probably enhance the map that you showed. Here is an example :

Here are some useful references on interferometry, aperture synthesis, cleaning, etc:
A Sightseeing Tour of mm Interferometry (Jérôme Pety)
Radio-astronomy Interferometry - Introduction (Philippe Salomé)
A mathematical approach to interferometry and cleaning:
Principles of image reconstruction in
Interferometry (Éric Thiébaut)
And here is a book recommendation: Tools of Radioastronomy
