Detecting planetary transits in the presence of stellar variability Optimal filtering and the use of colour information

Space-based photometric transit searches, such as the ESA mission Eddington (planned for launch in 2007), are expected to detect large numbers of terrestrial planets outside the solar system, including some Earth-like planets. Combining simplicity of concept with efficiency, the transit method consists in detecting the periodic luminosity drop in stellar light curves caused by the transit of a planet in front of its parent star. In a previous paper (Aigrain & Favata 2002, hereafter Paper I), we developed a Bayesian transit detection algorithm and evaluated its performance on simulated light curves dominated by photon noise. In this paper, we examine the impact of intrinsic stellar variability. Running the algorithm on light curves with added stellar variability (constructed using data from the VIRGO/PMO6 instrument on board SoHO) demonstrated the need for preprocessing to remove the stellar noise. We have developed an effective variability filter, based on an ad-hoc optimal approach, and run extensive simulations to test the filter and detection algorithm combination for a range of stellar magnitudes and activity levels. These show that activity levels up to solar maximum are not an obstacle to habitable planet detection. We also evaluated the benefits of using colour information in the detection process, and concluded that in the case of Eddington they are outweighed by the implied loss of photometric accuracy.