Skycorr: A general tool for spectroscopic sky subtraction

TL;DR

Skycorr is a versatile, instrument-independent tool that improves sky background subtraction in optical-to-near-IR spectroscopy by physically scaling airglow lines, significantly reducing residuals across various time scales.

Contribution

It introduces a novel, physically motivated method for sky subtraction that adaptively scales airglow lines in one-dimensional spectra, outperforming previous techniques.

Findings

Skycorr reduces sky residuals by several times compared to unscaled methods.

Performs well over time intervals from minutes to a year.

Outperforms the method of Davies (2007) in tests.

Abstract

Airglow emission lines, which dominate the optical-to-near-IR sky radiation, show strong, line-dependent variability on various time scales. Therefore, the subtraction of the sky background in the affected wavelength regime becomes a problem if plain sky spectra have to be taken at a different time as the astronomical data. A solution of this issue is the physically motivated scaling of the airglow lines in the plain sky data to fit the sky lines in the object spectrum. We have developed a corresponding instrument-independent approach based on one-dimensional spectra. Our code skycorr separates sky lines and sky/object continuum by an iterative approach involving a line finder and airglow line data. The sky lines are grouped according to their expected variability. The line groups in the sky data are then scaled to fit the sky in the science data. Required pixel-specific weights for overlapping groups are taken from a comprehensive airglow model. Deviations in the wavelength calibration are corrected by fitting Chebyshev polynomials and rebinning via asymmetric damped sinc kernels. The scaled sky lines and the sky continuum are subtracted separately. VLT X-Shooter data covering time intervals from two minutes to about one year were selected to illustrate the performance. Except for short time intervals of a few minutes, the sky line residuals were several times weaker than for sky subtraction without fitting. Further tests show that skycorr performs consistently better than the method of Davies (2007) developed for VLT SINFONI data.