Information on resource 'The HSOY Catalog'

HSOY is a catalog of 583'001'653 objects with precise astrometry based on PPMXL and Gaia DR1. Typical formal errors at mean epoch in proper motion are below 1 mas/yr for objects brighter than 10 mag, and about 5 mas/yr at the faint end (about 20 mag). South of -30 degrees, astrometry is significantly worse. HSOY also contains, where available, USNO-B, Gaia, and 2MASS photometry. HSOY's positions and proper motions are given for epoch J2000. The catalog becomes severely incomplete faintwards of 16 mag in the G-band. The mean epochs are typically very close to Gaia's J2015.

HSOY still contains about 0.7% spurious close "binaries" (non-matched stars) from the original USNO-B (marked with non-NULL clone). Also, failed matches within Gaia DR1 contribute another 1.5% spurious pairs (marked with non-NULL comp). In both cases, astrometry presumably is sub-standard.

More information is available at

The HSOY catalog is a common reduction of PPMXL and Gaia DR1. It essentially applies the technique discussed in 2010AJ....139.2440R, with PPMXL filling the role of USNO-B and Gaia DR1 filling the role of 2MASS. The concrete implementation is fairly different, though. Interested users are referred to the HSOY resource descriptor (this is a DaCHS file, but actual processing is written in SQL), in particular the add_gaia element.

Access Options

Like Gaia DR1, HSOY is primarily published through a TAP service; the primary site is ivo://org.gavo.dc/tap with the access URL

For simple tasks, there is an IVOA cone search service at (browser interface at

If you insist, the catalog content can also be downloaded in the form of an xz-compressed Postgres ASCII dump. This is a 40 GB download, and you'll probably spend more time making this useful for you than you'd spend learning ADQL and just using TAP. But if you insist or want to mirror HSOY: The data is (temporarily unavailable), the metadata (including the column sequence) at


The HSOY catalog contains 583'001'683 objects; this is significantly less than both the roughly 9e8 of PPMXL and the roughly 1.2e9 of Gaia DR1. The bulk of the objects missing in HSOY versus PPMXL should be non-stellar detections and failed matches PPMXL inherited from USNO-B1, but of course the inhomogeneous coverage of Gaia DR1 can be expected to play a role, too.

Positions are given for epoch J2000.0 in ICRS (as defined by Hipparcos via PPMX). Since Gaia astrometry is so precise, the mean epochs are fairly close to J2015 for almost all stars (mean: around 2014.8), even though many stars have observations going back to the 1950s (and older, for PPMXL's bright end is PPMX, which contains even older observations). This means that the positional errors given do not apply at J2000; formal errors in RA and Dec are of the order of 0.1 arcsec for most objects (tighter estimates can be derived fromt the errors in proper motion).

Since the Gaia position is so dominant, the mean error in position at mean epoch is of the order of the DR1 positional errors (about 3 mas averaged over the magnitude bins, much better at the bright end, much worse at the faint end).

Mean formal errors in proper motion at mean epoch range between better than 1 mas/yr for bright stars to about 5 mas/yr near the faint end. Note that these are really formal results of a simple least-squares reduction. They cannot be considered as absolute accuracies as PPMXL has spacial and magnitude dependent distortions in its proper motion system. They are slightly improved by incorporating Gaia observations, but they are not fully removed. Due to the short life expectancy of HSOY, a careful re-reduction of PPMXL is out of question.

This plot shows the the mean errors in proper motion in RA (red) and Dec (blue), in bins of 1 mag:


Since the early epochs are missing in USNO-B1 south of -30 degrees (no early surveys available), the errors in proper motion are much larger there; here is the distribution of errors in RA:


and in Dec:


To illustrate HSOY's coverage, this is a plot of the number of objects per level-6 healpix cell:


Here is a table of minimal, mean, and maximal values of the main HSOY columns:

col min avg max
raj2000 5.684374286825e-07 210.993388137 359.999993202
dej2000 -89.9928427252 -9.98338171978 89.9901000584
e_raepRA 4.74927e-09 9.14502574957e-07 2.38159e-05
e_deepDE 6.27509e-09 7.75452288327e-07 2.36765e-05
pmRA -0.00216075 -4.43904220127e-07 0.00242583
pmDE -0.00245665 -1.09799220875e-06 0.00257913
e_pmRA 1.14448e-08 8.34744285904e-07 2.68803e-06
e_pmDE 1.13905e-08 8.34238460892e-07 2.62362e-06
epRA 1975.62 2014.8336071 2015.0
epDE 1978.95 2014.88429072 2015.0
Jmag -0.676 15.1470283958 21.749
e_Jmag 0.013 0.0737119978739 9.999
Hmag -1.739 14.5851051481 24.292
e_Hmag 0.0 0.0891082665977 9.999
Kmag -2.099 14.3843849985 20.532
e_Kmag 0.0 0.102864557141 9.999
b1mag 1.6 18.6243196919 50.0
b2mag 0.014 18.7001648827 65.38
r1mag -3.159 17.3583684551 65.47
r2mag 0.024 17.710484873 62.38
imag 2.171 16.704232572 64.5
nobs 3 5.5826345577950630 17
gaia_id 65408 4045094535853177834 6917528993283204480
phot_g_mean_mag 3.15463 17.8191975575 28.6129
e_phot_g_mean_mag 1.06456e-05 0.00625220915035 0.109

G magnitudes and their errors are only given where the flux error as given in DR1 is smaller than 10% of the flux. In that case, the flux errors are translated into magnitude errors 1.09*flux_err/flux.

Known Flaws

(1) Duplicates: PPMXL contains many non-matched stars, i.e., two objects where only one should be present. Where the proper motion computed for these stars is roughly correct, these non-matched pairs end up in HSOY as well. All will be matched to the same Gaia DR1 object, which makes them behave rather odd kinematically (they "collide" at the epoch of Gaia). Such cases are marked with a 1 in the clone column. Clearly, astrometry has to be used with great care for these objects.

(2) Conversely, quite a few PPMXL objects match several Gaia DR1 objects. Since the true autocorrelation of Gaia DR1 (and several following releases) has a sharp drop at 2 arcsec for technical reasons, these objects do typically not correspond to multiple star systems but reflect match failures in the construction of Gaia DR1. In HSOY, such cases are marked with a non-NULL comp (which may remind you of component; in some rare cases, that association could actually match reality). This is a plot of (sum(comp)+sum(clone))/n in level-6 healpix cells, which is a good indicator of the severity of the problems with (1) and (2). The grid pattern represents the overlap areas in the surveys that entered into USNO-B, where problem (1) is particularly severe:


(3) As discussed in 2010AJ....139.2440R, PPMXL inherited from USNO-B about 2.5e7 spurious objects with high proper motions, mostly on the edges of plates, where pairs were matched that should not have been matched. Most of these objects did not hit Gaia DR1 objects when moved to J2015. Thus, HSOY only contains about 2.5e6 objects faster than 150 mas/yr (2.5e5 on the northern sky). Comparing this with the expectation of less than 1e5 from 2005AJ....129.1483L in the northern sky, it is clear that there still are many spurious high-PM objects in HSOY. To aid in discarding invalid high-PM objects, we matched PPMXL with SuperCOSMOS (cf. 2001MNRAS.326.1279H), which an independent extraction of the plates USNO-B is based on. Objects failing to match at J2000 received a 1 in HSOY's no_sc column. Note that for technical reasons, False in no_sc is encoded as NULL. Therefore, to filter out probably spurious objects, write WHERE no_sc IS NULL.

This purges another factor of two in bad high-PM objects all-sky (to 1.4e6). On the northern sky, the result is particularly pronounced, where only 1.7e5 objects with PM>150 mas/yr are left, well within a factor of two of the gold standard set by LSPM.

Here are comparisons of the distributions of total proper motions in HSOY for all (blue) and no_sc IS NULL (red) objects. Top or left is for the total catalog, bottom or right just the northern sky.

/hsoy/q/q/static/pm_histograms.png /hsoy/q/q/static/pm_histograms_north.png

To get an idea of where the obvious problem on the southern sky comes from, see this density map of stars faster than 150 mas/yr in HEALPixes of order 6, plotted in galactic coordinates. Only objects for which all of no_sc, comp, and clone are NULL are shown. Note that the scale on the aux axis is logarithmic:


Essentially, in areas so crowded that almost any movement of a star will hit another (within our match criteria) there will be many spurious high-pm stars, in particular when there are many spurious sources to begin with (plate borders). In sufficient distance from the bulge and the Magellanic clouds, high proper motions in HSOY should be fairly reliable when filtering out the suspicious (no_sc, clone, comp) objects.

(4) HSOY inherits the USNO photometry in B, R, and I. It has many known issues (including, but not limited to, ridiculous values like 65.47 shown in the table above). Do not lightly use b1mag, b2mag, r1mag, r2mag, and imag, as many of them are invalid.

(5) At the bright end, HSOY is severely incomplete. This reflects the corresponding incompleteness of Gaia DR1.


The primary key of HSOY is the tuple (ipix,comp). The suggested reference to a HSOY object is "HSOY <ipix>.0" when comp is NULL, and "HSOY <ipix>.<comp>" otherwise.

If you want to cite HSOY, please see our citation advice.

Services defined within this resource descriptor

Tables defined within this resource descriptor

Copyright, License, Acknowledgements

To the extent possible under law, the publisher has waived all copyright and related or neighboring rights to HSOY. For details, see the Creative Commons CC0 1.0 Public Domain dedication. Of course, you should still give proper credit when using this data as required by good scientific practice.


[Manage RD]