Assessing the performance of filtering algorithms for Earth-analogues observed with PLATO

Context: Detecting Earth-like planets presents a formidable challenge, primarily due to the influence of stellar activity, granulation and systematic errors on transit signals observed through photometric techniques. Stellar phenomena, including flares, spots, and convection, as well as the instrumental errors introduce complexities that obscure the identification of genuine planetary transits. Goal: This thesis aims to assess and compare the effectiveness of various filtering algorithms, with a specific focus on detecting Earth-like planets within the Habitable Zone, with orbital periods extending up to 2 years. The studied algorithms include the one developed by the research groups specifically to detrend the light curves from the stellar activity: Young Stars Detrending (YSD), as well as the all-purpose algorithms: biweight and Huber spline methods employed with 3 different window lengths (0.7, 1.4 and 2.0 days). The results hold particular relevance for the upcoming PLATO mission, as synthetic light curves were generated using the PLATO Solar-like Light-curve Simulator (PSLS). A series of injection-retrieval tests were conducted on these synthetic light curves to evaluate the performance of the selected filtering algorithms. Results: The biweight method and YSD Lowess regression emerge as the most effective algorithms for conducting a blind search for Earth-analogous planets. However, the precise retrieval of planetary parameters from the recovered transit signals remains challenging, as filtering algorithms distort the original signal. For the P1 sample representing the target stars during the initial two years of the mission, these algorithms fail to recover the planetary signal when applied to F5 spectral type stars. This is primarily due to the larger radii of such stars, which complicates detection by extending the period duration and reducing the planet-to-star radius ratio.