7.3.4 Quality assessment and validation

Author(s): László Molnár, Emese Plachy, Áron Juhász, Lorenzo Rimoldini

The verification of results and their validation are performed by employing:

  1. 1.

    SOS of Cepheids and RR Lyrae stars applied to sources with at least 12 FoV measurements in the G band (Section 7.4).

  2. 2.

    SOS of long period variables applied to sources with at least 12 FoV measurements in the G band (Section 7.7).

  3. 3.

    The crossmatch of Cepheids and RR Lyrae star candidates with objects in the Kepler/K2 fields (Section 7.3.4, Section 7.3.4).

  4. 4.

    Crossmatched objects not included in the training set (Rimoldini et al. 2018).


The verification of RR Lyrae and Cepheid candidates with Kepler/K2 fields is summarised here (for more details, see Molnár et al. 2018). We analysed the Gaia DR2 candidates in circular areas with a 8.5 degree radius centred on the fields of view of the original Kepler mission and the K2 mission observing Campaigns up to Field 13 (Howell et al. 2014, https://keplerscience.arc.nasa.gov/k2-fields.html). The prime Kepler mission observed a single field of view towards Lyra-Cygnus for four years. The K2 mission is ordered into campaigns along the Ecliptic; one campaign lasts for 60–80 days and then the spacecraft is reoriented. The Gaia DR2 candidates in these fields were crossmatched with the Kepler Input Catalog (KIC), the K2 Ecliptic Plane Input Catalog (EPIC), and the list of K2 targets selected for observation (Brown et al. 2011; Huber et al. 2016). The resolution of Kepler (4 ′′pixel-1) is much poorer than the one of Gaia, leading to some ambiguity in crossmatching the Gaia sources with the K2 targets. Nevertheless, RR Lyrae and Cepheid variations can be recovered even if the target is blended with another star within the photometric aperture of Kepler. We found no cases where two or more RR Lyrae or Cepheid candidates from Gaia would fall into the same aperture. We did not crossmatch sources from Campaign 9 that targeted the Galactic Bulge as the high source number density and the limited resolution of Kepler lead to strong confusion and data from OGLE was deemed superior to that of Kepler in this region.

We also made a list of known or suspected RR Lyrae stars that were proposed for observation and confirmed by Kepler, so that the completeness of Gaia DR2 candidates could be assessed from the rate of missed identifications.


The validation of RR Lyrae and Cepheid candidates with Kepler/K2 fields is summarised here (for more details, see Molnár et al. 2018). For the Lyra-Cygnus field, we visually inspected the Simple Aperture Photometry (SAP) and Pre-search Data Conditioning (PDC) SAP light curves of each target that was selected for observation in at least one observing quarter (one three-month segment of the original mission). We identified 48 RR Lyrae stars from the Gaia DR2 candidates, four of which were found not to be of the RR Lyrae type. Twelve other known RR Lyrae stars were not among the Gaia DR2 candidates, suggesting a sample completeness of about 78 per cent.

The original Kepler mission also acquired 52 Full-Frame Images (FFI). We extracted light curves for the objects not targeted by the mission from these images using the f3 code (Montet et al. 2017). We compared the light curves folded with the fundamental periods derived from the Gaia data as well as from the FFI data visually. Out of the 267 additional stars from the Gaia DR2 RR Lyrae candidates, we were able to classify 185 as RRAB or RRC variables (the other ones were either not RR Lyrae stars or associated with unreliable photometry). The combination of this set and the 48 stars described in the previous paragraph suggests a purity of the sample of at least 75 per cent.

In the K2 fields, we checked the light curves available for the targeted stars. These include the SAP/PDCSAP data sets provided by the mission as well as the available community-created light curves for selected campaigns. Out of the 1395 RR Lyrae candidates with counterparts in the K2 fields, 1371 were classified as RRAB or RRC in Gaia DR2, while 24 candidates turned out not to be RR Lyrae variables. The confirmed candidates are part of a larger set of 1816 known RR Lyrae stars in the K2 fields, suggesting a completeness rate around 75 per cent, in agreement with the one estimated from the original Kepler field, and a purity of 98 per cent (with a worst-case lower limit of 51 per cent) for the Ecliptic fields outside the Bulge. The interpretation of the purity value, however, is complicated by the biases in the selection of various targets for the K2 mission. About the classification of RR Lyrae stars into subclasses, 31 of the 1371 confirmed candidates were associated with the incorrect subtype, with misclassification rates of 1, 9, and 50 per cent for RRAB, RRC, and RRD types, respectively.

Cepheids were very sparse in the original Kepler fields. Among the Gaia DR2 Cepheid candidates, we found 38 Cepheid-type stars (ACEP, CEP, T2CEP) in the K2 fields and we were able to confirm 22, and assume 3 more of them (about 66 per cent). In the original field, we confirm the detection of the classical Cepheid V1154 Cyg and the T2CEP HP Lyr, while the semi regular star V677 Lyr was misclassified as T2CEP. However, the low number of targets prevented us from drawing more detailed conclusions.