Detrending algorithms in large time-series: Application to TFRM-PSES data

TL;DR

This paper demonstrates that combining TFA and Savitzky-Golay detrending algorithms significantly improves photometric data quality in large time-series, enabling better variability detection and faster processing through Python parallelization.

Contribution

The study applies and combines TFA and Savitzky-Golay algorithms to TFRM-PSES data, achieving substantial RMS reduction and enhanced computational efficiency.

Findings

Photometric RMS improved by a factor of 3-4 using combined methods.

Detection of more variable stars due to improved data quality.

Processing speed increased by a factor of ~10 with Python parallelization.

Abstract

Certain instrumental effects and data reduction anomalies introduce systematic errors in photometric time-series. Detrending algorithms such as the Trend Filtering Algorithm (TFA) (Kov\'{a}cs et al. 2004) have played a key role in minimizing the effects caused by these systematics. Here we present the results obtained after applying the TFA, Savitszky-Golay (Savitzky & Golay 1964) detrending algorithms and the Box Least Square phase folding algorithm (Kov\'{a}cs et al. 2002) to the TFRM-PSES data (Fors et al. 2013). Tests performed on this data show that by applying these two filtering methods together, the photometric RMS is on average improved by a factor of 3-4, with better efficiency towards brighter magnitudes, while applying TFA alone yields an improvement of a factor 1-2. As a result of this improvement, we are able to detect and analyze a large number of stars per TFRM-PSES field which present some kind of variability. Also, after porting these algorithms to Python and parallelizing them, we have improved, even for large data samples, the computing performance of the overall detrending+BLS algorithm by a factor of $\sim$10 with respect to Kov\'{a}cs et al. (2004).