MIARMA: A minimal-loss information method for filling gaps in time series

Abstract

The method here presented intends to minimize the effect of the gaps in the power spectra by gap-filling preserving the original information, that is, in the case of asteroseismology, the stellar oscillation frequency content. We make use of a forward-backward predictor based on autoregressive moving average modelling (ARMA) in the time domain. The method MIARMA is particularly suitable for replacing invalid data such as those present in the light curves of the CoRoT satellite due to the pass through the South Atlantic Anomaly, and eventually for the data gathered by the NASA planet hunter Kepler. We select a sample of stars from the ultra-precise photometry collected by the asteroseismic camera on board the CoRoT satellite: the {\delta} Scuti star HD 174966, showing periodic variations of the same order as the CoRoT observational window, the Be star HD 51193, showing longer time variations, and the solar-like HD 49933, with rapid time variations. We showed that in some cases linear interpolations are less reliable to what was believed. In particular: the power spectrum of HD 174966 is clearly aliased when this interpolation is used for filling the gaps; the light curve of HD 51193 presents a much more aliased spectrum than expected for a low frequency harmonic signal; and finally, although the linear interpolation does not affect noticeably the power spectrum of the CoRoT light curve of the solar-like star HD 49933, the ARMA interpolation showed rapid variations previously unidentified that ARMA interprets as a signal. In any case, the ARMA interpolation method provides a cleaner power spectrum, that is, less contaminated by spurious frequencies. In conclusion, MIARMA appears to be a suitable method for filling gaps in the light curves of pulsating stars observed by CoRoT since the method preserves their frequency content, which is a necessary condition for asteroseismic studies.