There have been many efforts to correct systematic effects in astronomical light curves to improve the detection and characterization of planetary transits and astrophysical variability in general. Algorithms like the Trend Filtering Algorithm (TFA) use simultaneously-observed stars to measure and remove systematic effects, and binning is used to reduce high-frequency random noise. We present TFAW, a modified version of TFA which reduces noise in variable-star light curves without modifying their intrinsic characteristics. We modified TFA’s iterative signal reconstruction by adding a Stationary Wavelet Transform filter which characterizes the noise- and trend-free signal and the underlying noise contribution at each iteration. The algorithm performs an adaptive noise estimation through the wavelet transform which reduces correlated and uncorrelated noise while preserving signals typical of astrophysical changes. We carried out tests over simulated sinusoidal and transit-like signals to assess the effectiveness of the method, and applied TFAW to real light curves from the Evryscope and TFRM. We also studied TFAW’s application to simulated multiperiodic signals. The TFAW improvement in RMS of simulated and real light curves ranges from 0.025 to 0.05 magnitudes compared to TFA. The signal-detection frequency power spectra remain almost unchanged for high SNR light curves, confirming that TFAW does not introduce new correlated noise sources. The signal detection efficiency of the power-spectrum peaks improves by a factor ~1.5 for low SNR light curves, allowing the recovery of transiting planets smaller than previous algorithms. TFAW is also able to improve the characterization of multiperiodic signals. We present two newly-discovered variable stars from Evryscope and TFRM. TFAW is a generic algorithm which is applicable to any kind of ground-based or space-based time-domain survey.
D. Ser, O. Fors, J. Nunez, et. al.
Thu, 23 Feb 17
Comments: Submitted to Astronomy & Astrophysics. 10 pages, 12 figures