Gaps in Gaia DR2 data

     

    A key element for a proper scientific interpretation of Gaia data is knowledge of various selection functions (see, for instance, Boubert et al. 2020). In turn, an essential part in establishing reliable selection functions is knowledge of the gaps that exist in the data, for instance those caused by spacecraft events such as refocussing, decontamination, or anomalies.

    This webpage provides tables with gaps in the data streams and data products of Gaia's second data release (Gaia DR2). Since astrometric, photometric, and spectroscopic data have been treated in different pipelines, separate gaps tables exist for each data type. The below tables exclusively apply to Gaia DR2; separate tables will be published for future data releases.

    Finally, it is important to acknowledge that gaps in the published data are also caused by the occasional on-board deletion of science data and that such gaps can show magnitude-dependent effects due to the priority management of the science data on board. These gaps are not covered here.

     

    Astrometry

    Section 2.1 of Lindegren et al. (2018) mentions that the Gaia DR2 astrometric processing has used data spanning revolutions 1192.13–3750.56, with major gaps covering revolutions 1316.49–1389.11 and 2324.90–2401.56 due to decontaminations. It is moreover mentioned that "planned maintenance operations (station-keeping manoeuvres, telescope refocusing, etc.), micrometeoroid hits, and other events caused additional gaps that rarely exceeded a few hours".

    The actual list of 239 astrometric gaps applicable to Gaia DR2 processing is available here. The definition of astrometric gaps applied here is somewhat arbitrary and, moreover, erroneously suggests that a black-and-white situation exists. In this light, one should recall that "bad" CCD observations have been downweighted in the Gaia DR2 astrometric solution through a statistical weight factor that depends on the size of the post-fit astrometric residual in relation to the expected uncertainty – see Equation (66) in Lindegren et al. (2012). If such downweighting is systematic over certain periods of time (and evidence exists that such situations do exist), it can also create gaps. As another example, a number of time intervals exist in which only six of the seven VPUs were generating science data, leading to partial gaps. Such (partial) gaps are not covered here.

     

    Photometry

    The photometry published in Gaia DR2 is far from a complete set. For instance, truncated windows have not been processed and neither have complex gate configurations. In addition, data has been removed from the streams due to, for instance, gaps in the offset non-uniformity libraries (affecting only the BP/RP processing) and/or failures in the offset non-uniformity determination and/or gaps in the background (straylight) correction maps. Moreover, various filters applied during the data processing have excluded data, for instance observations that have been flagged as problematic in the acquisition or image-parameter determination processes or outliers originating from cross-match problems (the latter have been removed by excluding any observation exhibiting a difference between the transit flux and the source reference flux exceeding 1 mag). Also, observations obtained with non-nominal configurations (e.g., gated faint-star windows) quite often have poor calibrations with a higher rate of failures when applied. Finally, a complex filtering scheme has been applied to select which sources to publish in the actual data release (see Arenou et al. 2018).

    Table 1 in Riello et al. (2018) furthermore mentions the existence of four photometric data gaps linked to two decontaminations and two refocussings. Riello et al. moreover mention that "outages in the daily processing pipelines [...] manifest [themselves] as minor gaps. These outages meant that some satellite telemetry was not actually available for Gaia DR2 processing, but [these gaps] will disappear in future date release[s]. Other gaps are instead caused by genuine spacecraft events and will never disappear".

    During the Gaia DR2 photometric processing, gaps in the data streams were noticed but the processing team merely performed a number of spot checks to cross-check the causes in order to ensure that these gaps had not been introduced by bugs or problems in the photometric processing. As a result, there is no documented list of photometric gaps for Gaia DR2. It seems safe, however, to assume to first order that photometric gaps are at least the same as astrometric gaps. This then results in a list of 10 photometric gaps, available here, composed of the four gaps from Riello et al. complemented with six astrometric gaps that fall in the EPSL period and that are not in the astrometric list discussed above. Added to these 10 gaps should be the 239 astrometric gaps discussed and provided above. There is (partial) overlap between these 239 astrometric gaps and the four photometric ones linked to the decontaminations and refocussings, which users should consider when merging these tables.

    Finally, it is important to stress that users who are concerned with selection functions also need to consider the effect of the changing magnitude limit of the survey resulting from the continuous build-up of contamination followed by "instantaneous" decontamination events. As can be seen in Figure 8 in Riello et al. (2018), this effect can be larger than 0.5 mag and is different between the three passbands, with BP affected most, followed by G, and RP affected least.

     

    Spectroscopy

    The median radial velocities published in Gaia DR2 are far from a complete set. For instance, spectra of several bright objects without astrometric coordinates have not been processed and neither have truncated windows nor spectra of faint objects (GRVS > 12 mag). In addition, stars with cool (≤3500 K) or hot (≥7000 K) cross-correlation template spectra, (suspected) SB2 binaries detected as double lined in more than 10% of their transits, emission-line stars, stars with ambiguous (variable) transit radial velocities, and median radial velocities for which the standard uncertainty exceeds 20 km s-1 have not been processed and/or published in Gaia DR2.

    Sartoretti et al. (2018) clarify that (the following quote has been slightly edited and updated for readability without changing the core of its contents) "spectra acquired during pre-defined time intervals, where the data are known to be of poor quality, have been filtered out. These intervals include the times during:

    • the decontaminations,
    • the refocussings,
    • the commissioning of the VPU version 2.8 software,
    • the time intervals where the astrometric AGIS residuals are high or where the AGIS solution did not produce a reconstructed spacecraft attitude (dubbed OGA-3), and 
    • the time gaps in the barycentric velocity correction (BVC) that are longer than 10 minutes.

    The total time covered by all the excluded intervals is ∼200 revolutions or ∼7.5% of the total observation time”.

    The actual list of 192 spectroscopic gaps applicable to Gaia DR2 is available here.

     

    Table description

    The description of the contents of each column is as follows:

    • Column 1 = start time of the gap, in revolutions as counted in the on-board mission timeline (OBMT). The relation between OBMT and calendar date and time is provided in the Gaia DR2 on-line documentation.
    • Column 2 = end time of the gap, in OBMT revolutions.
    • Column 3 = duration of the gap, computed as end time minus start time, in OBMT revolutions. One revolution is 21,600 s.
    • Column 4 = comment about the origin of the gap.
    • Column 5 = Boolean flag indicating whether the gap is persistent (true) or not (false). This column is only present for astrometric gaps. Examples of persistent gaps are those caused by decontaminations or micro-meteoroid hits. Examples of non-persistent gaps are those caused by delayed availability of (some) satellite telemetry during the Gaia DR2 processing. The interest in providing this flag is that some non-persistent gaps may manifest themselves as partial gaps, with some data still present.