2020-11-13T13:09:26Z
Geometric
and photogeometric distances to 1.47 billion stars in Gaia Early
Data Release 3 (eDR3)
[Note: UNPUBLISHED DRAFT. DO NOT USE JUST YET!]
We estimate the distance from the Sun of most sources in Gaia EDR3. We
provide two types of distance estimate, together with their
corresponding asymmetric uncertainties, using Bayesian posterior density
functions that we compute for every source. Our prior is based on a detailed
model of the spatial, velocity, colour, and magnitude distribution of stars in
our Galaxy that includes a 3D map of interstellar extinction.
The first type of distance estimate is purely geometric, in that it only makes
use of the Gaia parallax and parallax uncertainty. This uses a
line-of-sight-dependent distance prior derived from our Galaxy model. The
second type of distance estimate is photogeometric: in addition to parallax it
also uses the source's measured Gaia G magnitude and BP-RP colour. This type of
estimate uses the geometric prior together with a direction and
colour-dependent prior on the absolute magnitude of the star.
Our distance estimate and uncertainties are quantiles, so are invariant under
logarithmic transformations. This means that our median estimate of the
distance can be used to give the median estimate of the distance modulus.
.. _accompanying paper: http://www.mpia.de/homes/calj/gedr3_distances.html
milky-way-galaxy
stellar-distance
surveys
stars
Bailer-Jones, C.A.L.; et al. (TBD)
Gaia
[TODO: arXiv paper]
Research
Catalog
Optical
source_id
For each source we compute two posterior probability distributions over
distance: a geometric one and a photogeometric one. “Geometric” means
only parallax and the parallax uncertainty were used. “Photogeometric”
means the G magnitude and the BP-RP colour were used as well. For each
of these posterior distributions we estimate and provide three quantiles:
0.158655 (“Lo“), 0.5 (“Med”), and 0.841345 (“Hi”).
“Med” is the median of the distribution, and should be taken as the
distance estimate itself. “Lo” and “Hi” define the lower and upper ends
of the equal-tailed 68% (actually 68.269%) confidence interval this
estimate. If the posterior were Gaussian, then (r_hi_geo-r_lo_geo)/2
would be the 1-σ Gaussian uncertainty. However, we stress that these
confidence bounds are asymmetric, sometimes significantly so.
The distance estimates are predicated on the assumption that the source
is a single star in our Galaxy. Estimates are provided whereever
possible for sources that have the required input data, independent of
any other knowledge on the nature of that source (e.g. being a binary
star or quasar).
G_lim for that HEALpixel
:B:
The first (left-most) digit refers to the geometric posterior, the
second to the photogeometric posterior.
It indicates whether we have a low p-value (<1e-3) in Hartigan Dip test
(null hypothesis that posterior is unimodal, so small p suggests
evidence against this). Two-digit integer. Each digit can be:
:0:
not set, so assume unimodal hypothesis okay (or if test is not done
or gives no answer)
:1:
set, so possibly non-modality.
For instance, 10 means geo possibly multimodal, photgeo probably unimodal
Moreover, the computed confidence interval often spans any multimodality, so
generally speaking sources do not need to be excluded just because of
evidence of multimodality from this test.
:C:
QG models used to compute the photogeometric posterior. Two digit integer.
Each digit can be:
:0:
NULL
:1:
one-component Gaussian model
:2:
two-component Gaussian model
:3:
spline
The first (left-most) digit refers to the lower (bluer) model, the second to
the upper (redder) model.
E.g. 13 means the lower one was a one-component Gaussian and the upper
one was was spline.
There are also two special setting of this flag:
:99:
data (G or BP-RP) were missing (so no photogeo distance could be computed)
:88:
no attempt was made to compute photogeo
TODO: 88 may never occur in practice. If it does not, we should remove
this.
]]>
0/0-11
cone/scs.xml
```
# The actual assertions are pyUnit-like. Obviously, you want to
# remove the print statement once you've worked out what to test
# against.
row = self.getFirstVOTableRow()
print(row)
self.assertAlmostEqual(row["ra"], 22.22222)
```