Automated Identification of the Most Likely Source of a Transit Signal in Kepler and TESS Data

Background eclipsing binaries (BGEBs) are a dominant source of false positives in transit surveys, for example accounting for as much as 40% of Kepler transit detections in crowded fields at low Galactic latitudes. BGEBs are particularly prevalent in TESS data due to the larger TESS pixels and observations at lower Galactic latitudes than Kepler. Several techniques to identify BGEB false positives were developed for analysis of Kepler data, but the performance of these methods degraded with decreasing SNR. Automated methods developed for uniform Kepler vetting, required to support occurrence rate estimates, only gave a yes/no determination, with little information about the confidence of that determination.

We describe a method that, for each star near the pixels used to create photometric light curves for transit detection, computes the probability that this star is the source of the transit signal relative to other nearby stars. Our method models transits on each nearby star and computes a Bayesian probability that a particular star is the transit source relative to other nearby stars. This method improves on similar techniques developed for Kepler data by a) reducing the reliance on PSF fitting, which fails at low SNR and b) modeling transits using flight data, in particular actual noise characteristics rather than simple noise models. We demonstrate our methods on Kepler and TESS data, showing that this method is more robust at low SNR than the previous method developed for Kepler, and therefore is suited to automated vetting. We will present a catalog giving the most likely transit source for Kepler and TESS planet candidates.

Our method enables science in several ways: a) effective high-reliability information about which star hosts a transit can significantly reduce the number of false positives that are followed up, reducing follow-up costs, b) provide improved identification of exoplanet host stars enabling more accurate exoplanet characterization, and c) is an important step towards using TESS data for exoplanet occurrence rate studies. Automated vetting of TESS data is critical for such studies.

This work is supported by TESS Guest Investigator grant #20-TESS20-0042.