Quantifying energy fluence and its uncertainty for radio emission from particle cascades in the presence of noise

TL;DR

This paper introduces a frequency-domain method to accurately estimate the energy fluence of radio signals from particle cascades, reducing bias and uncertainty issues caused by noise superposition, especially at low SNR.

Contribution

It presents a novel frequency-domain approach for quantifying radio emission energy fluence, addressing biases and uncertainties inherent in traditional time-domain noise subtraction methods.

Findings

Reduces bias in energy fluence estimation at low SNR

Provides a more accurate uncertainty quantification method

Enables frequency-dependent fluence analysis

Abstract

Measurements of radio signals induced by an astroparticle generating a cascade present a challenge because they are always superposed with an irreducible noise contribution. Quantifying these signals constitutes a non-trivial task, especially at low signal-to-noise ratios (SNR). Because of the randomness of the noise phase, the measurements can be either a constructive or a destructive superposition of signal and noise. To recover the electromagnetic energy of the cascade from the radio measurements, the energy fluence, i.e. the time integral of the Poynting vector, has to be estimated. Conventionally, noise subtraction in the time domain has been employed for energy fluence reconstruction, yielding significant biases, including even non-physical and negative values. To mitigate the effect of this bias, usually an SNR threshold cut is imposed, at the expense of excluding valuable data from the analyses. Additionally, the uncertainties derived from the conventional method are underestimated, even for large SNR values. This work tackles these challenges by detailing a method to correctly estimate the uncertainties and lower the reconstruction bias in quantifying radio signals, thereby, ideally, eliminating the need for an SNR cut. The development of the method is based on a robust theoretical and statistical background, and the estimation of the fluence is performed in the frequency domain, allowing for the improvement of further analyses by providing access to frequency-dependent fluence estimation.