Bayesian Methods for Joint Exoplanet Transit Detection and Systematic Noise Characterization

Abstract

Abstract The treatment of systematic noise is a significant aspect of transit exoplanet data processing due to the signal strength of systematic noise relative to a transit signal. Typically, the standard approach to transit detection is to estimate and remove systematic noise independently of and prior to a transit detection test. If a transit signal is present in a light curve, the process of systematic noise removal may distort the transit signal by overfitting and thereby reduce detection efficiency. We present a Bayesian framework for joint detection of transit signals and systematic noise characterization and describe the implementation of these detectors as optimal Neyman–Pearson likelihood ratio tests. The joint detectors reduce to closed form as matched filters under the assumption of a Gaussian Bayesian prior for the systematic noise. The performance of the exploratory detectors was evaluated in injection tests and show ∼2% improvement in overall detection efficiency relative to the standard approach. We find that joint detection efficiency is specifically improved for short-period, low transit-depth exoplanet transits, providing evidence in support of the hypothesis that joint detection may indeed help to mitigate overfitting. In addition, an initial feasibility test to detect known exoplanets in Kepler data using the joint detectors produced encouraging preliminary results.