MWS SpecDis Catalog¶
Overview¶
We present our SpecDis value added stellar distance catalog accompanying the Dark Energy Spectroscopic Instrument (DESI) survey Data Release 1. SpecDis involves training a feed-forward multilayer perceptron Neural Network (NN) on a large sample of stars with Gaia parallaxes, but without applying any selection on either parallax error or signal-to-noise (S/N) ratio of the stellar spectra. Instead we incorporate the Gaia parallax measurement error into the loss function for the training. This approach ensures that the training sample does not suffer from biases in parallax. To enhance the precision of distance predictions, we employ Principal Component Analysis to reduce noise and dimensionality of the input stellar spectra. Validated by an independent external sample of member stars with precise distance measurements from globular clusters, dwarf galaxies, and stellar streams, we demonstrate that our distance measurements show no significant bias up to 100 kpc, and are significantly more precise than Gaia parallax beyond 7 kpc. The median distance uncertainties are 23% for S/N<20, 19% for 20<=S/N<60, 11% for 60<=S/N<100, and 7% for S/N>=100. Additionally, we develop a Gaussian mixture model to identify candidate binary systems by modeling the discrepancy between the NN-predicted absolute magnitudes and the geometric absolute magnitudes derived from Gaia G-band apparent magnitude and parallaxes. With this model, we have identified 120000 possible binaries. Our final catalog provides distance and distance uncertainty measurements for over 4 million stars, offering a valuable resource for Galactic astronomy. There are a total of 4140838 stars in this distance catalog, with 101299 stars having log(g) smaller than 3. More details about SpecDis can be found in Li et al., 2025.
The DESI Milky Way Survey provides two stellar distance Value Added Catalogs: SpecDis (this catalog) and SPdist. The major difference between SpecDis and SPdist is that SpecDis predicts distances from the stellar spectra, whereas SPdist predicts distances from a list of stellar parameters by the DESI Milky Way Survey pipelines.
Data Access¶
Data URL: https://data.desi.lbl.gov/public/dr1/vac/dr1/mws-specdis
NERSC access:
/global/cfs/cdirs/desi/public/dr1/vac/dr1/mws-specdis
Documentation¶
Files¶
iron-yr1-v2.0.fits: Distance catalog of DESI Data Release 1
Data Model¶
| Name | Type | Units | Description |
|---|---|---|---|
| TARGETID | int64 | – | DESI source ID |
| SOURCE_ID | int64 | – | Gaia DR3 source ID |
| RA | float64 | deg | Gaia DR3 Right Ascension |
| DEC | float64 | deg | Gaia DR3 Declination |
| PMRA | float64 | mas / yr | Gaia DR3 Proper Motion in Right Ascension |
| PMRA_ERR | float64 | mas / yr | Uncertainty in pmra |
| PMDEC | float64 | mas / yr | Gaia DR3 Proper Motion in Declination |
| PMDEC_ERR | float64 | mas / yr | Uncertainty in pmdec |
| VRAD | float64 | km/s | Radial velocity |
| DIST | float64 | kpc | Heliocentric distances |
| DISTERR | float64 | kpc | Uncertainty of distance |
| MG_NN | float64 | – | NN predicted Gaia G-band absolute magnitude |
| MG_GEO | float64 | – | Observed Gaia G-band absolute magnitude |
| PARALLAX | float64 | mas | Gaia DR3 parallax before zero point correction |
| PARALLAX_ERR | float64 | mas | Uncertainty in parallax |
| PARALLAX_ZPC | float64 | mas | Zero point correction of parallax |
| EBV | float64 | – | Reddening estimated in this work |
| A_G | float64 | – | Dust correction in Gaia G-band |
| RUWE | float64 | – | Gaia DR3 RUWE |
| APS | float64 | – | Gaia DR3 ASTROMETRIC_PARAMS_SOLVED |
| NEUIA | float64 | – | Gaia DR3 NU_EFF_USED_IN_ASTROMETRY |
| P_COLOUR | float64 | – | Gaia DR3 PSEUDOCOLOUR |
| ECL_LAT | float64 | – | Gaia DR3 ECL_LAT |
| BINARY_FLAG | int64 | – | Flag of binaries: 1 for single stars, 0 for binaries |
| BINARY_POSSIBILITY | float64 | – | Binary possibility of a star |
Contact¶
Contact Songting Li with any questions.