Interpolation of Instrument Response Functions for the Cherenkov Telescope Array in the Context of pyirf

Abstract

The Cherenkov Telescope Array (CTA) will be the next generation ground-based very-high-energy gamma-ray observatory, constituted by tens of Imaging Atmospheric Cherenkov Telescopes at two sites once its construction and commissioning are finished. Like its predecessors, CTA relies on Instrument Response Functions (IRFs) to relate the observed and reconstructed properties to the true ones of the primary gamma-ray photons. IRFs are needed for the proper reconstruction of spectral and spatial information of the observed sources and are thus among the data products issued to the observatory users. They are derived from Monte Carlo simulations, depend on observation conditions like the telescope pointing direction or the atmospheric transparency and can evolve with time as hardware ages or is replaced. Producing a complete set of IRFs from simulations for every observation taken is a time-consuming task and not feasible when releasing data products on short timescales. Consequently, interpolation techniques on simulated IRFs are investigated to quickly estimate IRFs for specific observation conditions. However, as some of the IRFs constituents are given as probability distributions, specialized methods are needed. This contribution summarizes and compares the feasibility of multiple approaches to interpolate IRF components in the context of the pyirf python software package and IRFs simulated for the Large-Sized Telescope prototype (LST-1). We will also give an overview of the current functionalities implemented in pyirf.

ICRC 2023 Rapporteur: Status of the Gamma-ray Astronomy field

“…The software development is also essential for the advancement of Science, in particular in fields in which the analysis of Big Data is a must. There were several presentations about methods, simulations and analysis software like ctapipe [196], the low-level data processing pipeline software for CTAO, gammapy [197] the high level one, pyirf [198] the software to build the instrument response functions in CTAO, astripipe [199] the ASTRI pipeline, NectarChain [200] Nectar one or cosipy [201], the COSI pipeline or libraries to fit data (Bjet_MCMC) [202] or catalogs (STeVECat [203]). It is also worth noting the role that deep learning is starting to acquire, not only to improve the data processing, but also in any task related to population classification or regression of some quantity, for which we had more than ten contributions in the conference.…”