Advanced modelling of Night Sky Background light for Imaging Atmospheric Cherenkov Telescopes

TL;DR

This paper presents a physics-based, wavelength-dependent model for Night Sky Background light in Imaging Atmospheric Cherenkov Telescopes, improving prediction accuracy over standard constant-background models.

Contribution

The authors developed a detailed, physics-driven simulation of NSB for IACTs, incorporating atmospheric and telescope-specific factors, validated against H.E.S.S. data, and made the software publicly available.

Findings

Per-Pixel predictions match H.E.S.S. data within ±21-19%.

The model significantly outperforms constant-background assumptions.

Validation shows small deviations due to atmospheric variability.

Abstract

A significant source of noise for Imaging Atmospheric Cherenkov Telescopes (IACTs), which are designed to measure air showers caused by astrophysical gamma rays, is optical light emitted from the night sky. This Night Sky Background (NSB) influences IACT operating times and their sensitivity. Thus, for scheduling observations and instrument simulation, an accurate estimate of the NSB is important. A physics-driven approach to simulating wavelength-dependent, per-photomultiplier-pixel NSB was developed. It includes contributions from scattered moonlight, starlight, diffuse galactic light, zodiacal light, and airglow emission. It also accounts for the absorption and scattering of optical light in the atmosphere and telescope-specific factors such as mirror reflectivity, photon detection efficiency, and focal length. The simulated results are corrected for pointing inaccuracies and individual pixel sensitivities and compared to data from the High Energy Stereoscopic System (H.E.S.S.) IACT array. The software package developed for this analysis will be made publicly available. Validation against H.E.S.S.\ data shows small deviations from the prediction, attributable to airglow and atmospheric variability. Per-Pixel predictions provide a good match to the data, with the relative 90\% error range being [-21\%, 19\%]. Compared to the existing standard modelling approach of assuming a constant background, where the relative 90\% error range was [-64\%, 48\%], this is a significant improvement.