skip to main content

gaia early data release 3 documentation

Gaia Early Data Release 3
Documentation release 1.0

European Space Agency (ESA)
Gaia Data Processing and Analysis Consortium (DPAC)
3 December 2020
Executive summary

The early installment of the third Gaia data release, Gaia EDR3, encompasses astrometry and photometry, complemented with radial velocities copied from Gaia DR2 after removal of a small number of spurious entries. The basic number statistics of the contents of Gaia EDR3 is as follows:

Data product or source type Number of sources
Total 1 811 709 771
Five-parameter astrometry (position, parallax, proper motion) 585 416 709
Six-parameter astrometry (position, parallax, proper motion, pseudo-colour) 882 328 109
Two-parameter astrometry (position only) 343 964 953
Gaia-CRF3 extra-galactic sources (optical reference frame) 1 614 173
ICRF3 sources (for frame orientation) 2 269
Gaia-CRF3 sources (for frame spin) 429 249
G-band (330–1050 nm) 1 806 254 432
GBP-band (330–680 nm) 1 542 033 472
GRP-band (630–1050 nm) 1 554 997 939
Median radial velocity over 22 months (Gaia DR2) 7 209 831

The basic quality statistics of the contents of Gaia EDR3 are as follows (where the astrometric uncertainties as well as the Gaia-CRF3 alignment and rotation (spin) limits refer to epoch J2016.0 TCB):

Data product or source type Typical uncertainty
Five-parameter astrometry (position) 0.010.02 mas at G<15
0.05 mas at G=17
0.4 mas at G=20
1.0 mas at G=21
Five-parameter astrometry (parallax) 0.020.03 mas at G<15
0.07 mas at G=17
0.5 mas at G=20
1.3 mas at G=21
Five-parameter astrometry (proper motion) 0.020.03 mas yr-1 at G<15
0.07 mas yr-1 at G=17
0.5 mas yr-1 at G=20
1.4 mas yr-1 at G=21
Six-parameter astrometry (position) 0.020.03 mas at G<15
0.08 mas at G=17
0.4 mas at G=20
1.0 mas at G=21
Six-parameter astrometry (parallax) 0.020.04 mas at G<15
0.1 mas at G=17
0.5 mas at G=20
1.4 mas at G=21
Six-parameter astrometry (proper motion) 0.020.04 mas yr-1 at G<15
0.1 mas yr-1 at G=17
0.6 mas yr-1 at G=20
1.5 mas yr-1 at G=21
Two-parameter astrometry (position only) 13 mas
Systematic astrometric errors (sky averaged) <0.05 mas
Gaia-CRF3 alignment with ICRF 0.01 mas at G=19
Gaia-CRF3 rotation with respect to ICRF <0.01 mas yr-1 at G=19
Mean G-band photometry 0.3 mmag at G<13
1 mmag at G=17
6 mmag at G=20
Mean GBP-band photometry 0.9 mmag at G<13
12 mmag at G=17
108 mmag at G=20
Mean GRP-band photometry 0.6 mmag at G<13
6 mmag at G=17
52 mmag at G=20
Median radial velocity over 22 months 0.3 km s-1 at GRVS<8
0.6 km s-1 at GRVS=10
1.8 km s-1 at GRVS=11.75
Systematic radial velocity errors <0.1 km s-1 at GRVS<9
0.5 km s-1 at GRVS=11.75

The Gaia EDR3 parallaxes show evidence for a global zero point, in the sense Gaia - ‘true’, of about -0.017 mas, which has not been ‘corrected’ in the data (for details, see Lindegren et al. 2020).

The data collected between 25 July 2014 and 28 May 2017 – during the first 34 months of the nominal, five-year mission – have been processed by the Gaia Data Processing and Analysis Consortium (DPAC), resulting into this third data release. A summary of the release properties is provided in Gaia Collaboration et al. (2020b). The overall scientific validation of the data is described in Fabricius et al. (2020). These papers are considered ‘must-read’ material for any user of Gaia EDR3 data. Background information on the Gaia mission and the spacecraft can be found in Gaia Collaboration et al. (2016b), with a more detailed presentation of the Radial Velocity Spectrometer (RVS) in Cropper et al. (2018). In addition, Gaia EDR3 is accompanied by dedicated papers, all part of a Special Issue of A&A, that describe the processing and validation of the various data products: Lindegren et al. (2020) for the Gaia EDR3 astrometry, Riello et al. (2020) for the Gaia EDR3 photometry, Seabroke et al. (2020) for the radial velocities published in Gaia EDR3, and Klioner et al. (2020) for the celestial reference frame. Four more papers present a glimpse of the scientific value, richness, and potential of the data in the areas of the kinematics of the Milky Way towards the Galactic anti-centre (Gaia Collaboration et al. 2020a), the properties of the complete sample of nearby stars (Gaia Collaboration et al. 2020c, , the Gaia Catalogue of Nearby Stars), the structure and properties of the Magellanic Clouds (Gaia Collaboration et al. 2020b), and the acceleration of the solar-system barycentre with respect to distant quasars (Gaia Collaboration et al. 2020a). In addition to the set of references mentioned above, this documentation provides a detailed, complete overview of the processing and validation of the Gaia EDR3 data.

Data from and Gaia EDR3, as well as from Gaia DR1 and Gaia DR2, can be retrieved from the Gaia ESA Archive (GEA), which is accessible from The archive also provides various tutorials on data access and data queries plus an integrated data model (i.e., description of the various fields in the data tables). In addition, Luri et al. (2018) provide concrete advice on how to deal with Gaia astrometry, with recommendations on how best to estimate distances from parallaxes. The Gaia archive features a visualisation service which can be used for quick initial explorations of the entire Gaia EDR3 data set. Carefully validated, pre-computed cross matches between Gaia EDR3 and a selected of large surveys is provided, with details described in Marrese et al. (2019, 2020). Finally, Gaia EDR3 contains the (intended) pointing of the Gaia telescopes as a function of time (commanded_scan_law table) and simulated Gaia catalogues (gaia_universe_model and gaia_source_simulation tables).

Gaia EDR3 represents a major advance with respect to Gaia DR2 in terms of astrometric and photometric precision and accuracy as well as in survey homogeneity across colour, magnitude, and celestial position. For the first time, the astrometric solution has benefited from an iterative step between the determination of image locations and fluxes and the astrometric calibrations. Further notable improvements include:

  • Enhancements in the creation of the source list that make it more robust with respect to variable stars, high proper motion stars, and the disturbing effects of spurious on-board detections – caused, among others, by diffraction spikes of bright stars, galactic cosmic rays, or planets in the solar system transiting the fied of view – and partially resolved sources (Torra et al. 2020);

  • A more sophisticated modelling, including temporal and source-colour dependencies, of the line and point spread functions of the astrometric instrument in the image parameter determination (Rowell et al. 2020);

  • An improved and extended astrometric calibration model, to better handle saturated images and to cope with the effects of radiation-induced charge transfer inefficiency, imperfections in the point and line spread function models that cause residual effects at the sub-pixel level, and the variable rate with which sources move across the focal plane in the direction perpendicular to the scan direction;

  • An improved and extended photometric calibration model, leading to a better treatment of saturated images and a more precise and fine-grained determination of the background flux – either due to stray light or due to astronomical sources – in each observation window. Gaia EDR3 also uses an improved set of external photometric calibrators that are more evenly distributed in colour and in magnitude than for Gaia DR2. As a consequence, and in contrast to Gaia DR2, a single passband for each of the photometric bands G, GBP, and GRP can now be used over the entire magnitude and colour range, with systematics remaining below 1%.

Gaia EDR3 provides an updated materialisation of the celestial reference frame at optical wavelengths. The so-called Gaia-CRF3, which is based solely on extragalactic sources, is aligned with the International Celestial Reference Frame (ICRF) and details are provided in Klioner et al. (2020). An ad-hoc correction has been introduced to ensure that the bright-star reference frame has no net spin with respect to the reference frame defined by (faint) quasars. The tables agn_cross_id and frame_rotator_source provide the source IDs of the Gaia-CRF3 sources.

Several limitations and best practices for the use of the Gaia EDR3 data exist (as described in Gaia Collaboration et al. 2020b) and we summarise here the most important ones that users of Gaia EDR3 should be aware of:

  • The validation of Gaia EDR3 was done in various stages and led to the decision to filter out small parts of the available data before publication (see also Fabricius et al. 2020). The applied filters are summarised in Gaia Collaboration et al. (2020b) and details can be found in the data processing papers listed above and in this documentation.

  • The survey represented by Gaia EDR3 is essentially complete between G=12 and G=17 mag. At the bright end (G<7 mag), the completeness has essentially remained unchanged compared to Gaia DR2 while it has slightly improved at the faint end. More details on the survey completeness can be found in the papers listed above and in Chapter 7. No attempt has been made at deriving a selection function for Gaia EDR3.

  • During the data processing for this release (and all earlier releases), all sources have been treated as single stars. This means that, for binary and multiple stellar systems, the astrometry – as well as the median radial velocity – is less accurate since the astrometric parameters may refer to either component or to the photocentre of the system such that, for instance, the proper motion may represent the mean motion of either component or of the photocentre over the 34 months of data included in the Gaia EDR3 solution. Sources that are not single stars are not marked as such in the catalogue.

  • In the astrometric processing, the colour of each source has been included in the form of the effective wave number. This wave number has preferentially been derived from the flux as a function of wavelength in the BP and RP prism spectra, leading to classical 5-parameter astrometric solutions, or has been estimated as pseudo-colour along with the astrometric parameters in the astrometric processing, leading to 6-parameter astrometric solutions. The resulting two solution types have globally different uncertainties and systematics, with 6-parameter astrometry in general being less precise. For 5-parameter solutions, the published astrometric uncertainties are underestimated by 5% at the faint end (G>16 mag) and by up to 30% at the bright end (G<14 mag). For 6-parameter solutions, these numbers are 20% and up to 40%, respectively. As shown in Lindegren et al. (2020), the underestimation of the published uncertainties increases in crowded areas, such as the Magellanic Clouds, and for sources that have indications that they may have companions or be part of a partially resolved double.

  • For sources that are separated on the sky by 0.20.3 arcsec, such that they are only occasionally resolved in the Gaia transits, ambiguity in the observation-to-source matching can lead to spurious parallax values that are very large (positive or negative) and appear highly significant. Such sources tend to be faint and located in crowded regions. Guidance on how to clean samples from spurious astrometry based on the renormalised unit weight error (RUWE) diagnostic, that is part of Gaia EDR3, is provided in Lindegren et al. (2020).

  • The spatial resolution of Gaia EDR3 has improved with respect to Gaia DR2 and incompleteness in close pairs of stars starts below separations of 1.5 arcsec. Below 0.7 arcsec, the completeness in close source pairs decreases very rapidly. Nonetheless, the treatment of such sources has been improved and close pairs with separations between 0.18 and 0.4 arcsec which were erroneously considered duplicate sources in Gaia DR2 appear as two sources in Gaia EDR3 (although such pairs may still represent spurious solutions). New quality indicators in Gaia EDR3 that are related to the image parameter determination step provide useful indications whether, for instance, a source is one of a close pair (and possibly a binary) or whether it suffers from nearby disturbing sources. Fabricius et al. (2020) shows, for instance, that the parameter ipd_gof_harmonic_amplitude is useful for identifying spurious solutions of resolved doubles, which are not correctly handled in the Gaia EDR3 astrometric processing.

  • Systematic errors in the parallaxes are estimated to be below the 0.05 mas level (Lindegren et al. 2020). As in Gaia DR2, there is an overall parallax zero point which, from an examination of QSO parallaxes, is estimated to be -0.017 mas (in the sense of the Gaia EDR3 parallaxes being too small). The parallax zero point depends on the sample of sources examined and varies as a function of magnitude, colour, and celestial position. A tentative correction recipe, treating sources with 5- and 6-parameter astrometry separately, to remove the parallax bias as a function of source magnitude, colour, and ecliptic latitude is presented in Lindegren et al. (2020). In addition to a global zero point, there are also regional systematic errors as well as source-to-source correlations in the errors (which also affect the other astrometric parameters). More information is provided in Lindegren et al. (2020), Lindegren et al. (2020), Fabricius et al. (2020), and Chapter 7.

  • Due to background estimation inaccuracies, source colours for faint stars, in crowded regions, and in the surroundings of bright stars can be unreliable. One should treat colour-magnitude diagrams constructed for these cases with care. For faint red sources, the flux in the BP band is typically overestimated which causes such sources to appear much bluer in (GBP-GRP) than they should be. When studying the lower main sequence, using (G-GRP) as colour index is recommended.

  • The Gaia EDR3 catalogue contains a field, phot_bp_rp_excess_factor, which can be used to judge the extent to which the photometry of a given source is compromised. Compared to Gaia DR2, the flux excess factor in Gaia EDR3 is more representative of astrophysical inconsistencies between the fluxes in BP/RP and G, for example due to the extended nature of a source or a non-standard (non-stellar) spectral energy distribution. In contrast to Gaia DR2, no filtering on the flux excess factor has been done in Gaia EDR3. Users are recommended to apply a suitable filter themselves, following the guidelines defined in Riello et al. (2020). This paper also presents a corrected version of the flux excess factor which is recommended for use instead of the ‘raw’ phot_bp_rp_excess_factor value published in Gaia EDR3. Two new photometric data quality indicators are included in Gaia EDR3 (phot_bp_n_blended_transits and phot_bp_n_contaminated_transits, and similar for RP). These allow filtering of sources according to the probability that their photometry is affected by crowding. Details can be found in Riello et al. (2020) and in Chapter 7 of this documentation.

  • The published magnitudes of bright stars should be corrected for saturation effects (bright means G<8 mag for G photometry and G<4 mag for GBP and GRP photometry). Users are also advised to correct the published G-band photometry for sources fainter than G=13 mag with 6-parameter astrometric solutions (field astrometric_params_solved=95) to bring them onto the photometric system of the 5-parameter sources. Details are provided in Riello et al. (2020).

  • Gaia EDR3 does not contain new radial velocities but essentially contains copies of the Gaia DR2 values (see Katz et al. 2019)). Details are explained in Seabroke et al. (2020), where it is clarified that a careful tracing of Gaia DR2 sources to their Gaia EDR3 counterparts and a filtering of spurious values (mostly in the extreme tails of the distribution) has been implemented, resulting in 7 209 831 entries (with 97% of the Gaia EDR3 sources with a radial velocity having the same source ID as in Gaia DR2).

  • Gaia EDR3 should be treated as an independent catalogue from Gaia DR2. In particular the photometric systems of the two catalogues differ (see Riello et al. 2020) and the source list has changed (see Torra et al. 2020). Although changes in the source list are modest (affecting less than 3% of the sources), tracing sources from Gaia DR2 to Gaia EDR3 should not be done by blindly matching source IDs but through us of the dr2_neighbourhood match table from Gaia DR2 to Gaia EDR3 (see Marrese et al. 2020, and Section 12.3).

Summary of miscellaneous links:

List of Figures:
List of Tables: