🔭 astrophysics

SPORE: An Event-Level Sampling Pipeline for Multi-Telescope Neutrino Astronomy

This paper introduces SPORE, an open-source Python package that simulates multi-telescope neutrino events using tabulated instrument response functions, validating its accuracy and flexibility through comparisons with public IceCube data releases.

Original authors: Jeffrey Lazar, Perrine Wilmet, Gwenhaël de Wasseige

Published 2026-08-13

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Jeffrey Lazar, Perrine Wilmet, Gwenhaël de Wasseige

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, chaotic party where most guests are light-based—photons bouncing off stars, galaxies, and gas clouds. But sometimes, the party throws a different kind of guest: a neutrino. These are ghostly, tiny particles that barely interact with anything. They can zip through entire planets without bumping into a single atom, carrying secret messages from the most violent, hidden corners of the cosmos, like the hearts of exploding stars or the jets of black holes. Because they don't get blocked by dust or gas, they are the only messengers that can tell us what's happening in places where light gets trapped.

To catch these ghosts, scientists build massive detectors deep under the ice or in the ocean, waiting for a rare collision that creates a flash of blue light. But catching a ghost is hard; you need to know exactly how your detector "sees" the world. This involves three tricky things: how big a net the detector casts (effective area), how blurry its vision is (point spread function), and how well it guesses the energy of the particle (energy resolution). Before this paper, if you wanted to simulate what these detectors would see, you either had to run incredibly slow, complex computer simulations from scratch or use tools that only gave you rough averages, not individual events. It was like trying to predict the outcome of a dice game by only knowing the average roll, rather than rolling the dice yourself.

Enter spore, a new open-source tool created by researchers Jeffrey Lazar, Perrine Wilmet, and Gwenhaël de Wasseige. Think of spore as a "ghost-hunting simulator" that lets scientists generate realistic, individual neutrino events without needing the heavy-duty, proprietary software usually required. Instead of building a detector from the ground up, spore takes a "response file"—a pre-made map of how a specific telescope sees the universe—and uses it to roll the dice for millions of fake events. It can simulate a single point of light in the sky, a fuzzy extended glow, or even a team of different telescopes working together.

The team tested spore by feeding it the real data maps from the famous IceCube detector in Antarctica. They asked spore to generate a fake set of neutrino events and then compared the results to what IceCube actually saw. The results were impressive: spore successfully recreated the distribution of where the neutrinos came from in the sky, matching the real data within about 10–15%. It also did a good job matching the energy of the events for most of the sample. However, the team found a small hiccup: for very low-energy events (below 600 GeV), spore predicted more events than were actually observed. They traced this not to a flaw in their simulator, but to the "blurriness" of the original data maps they used; the maps were too coarse to handle the low-energy details perfectly.

In short, spore doesn't claim to have discovered a new particle or solved the mystery of the universe. Instead, it provides a vital, flexible, and fast way for scientists to test their ideas. It allows them to ask, "If our theory is right, what would our telescopes see?" and get a realistic answer instantly, whether they are looking at a single star or a whole galaxy, and whether they are using one telescope or a global team of them. It fills a gap in the toolkit, making it easier for the next generation of neutrino astronomers to explore the dark, ghostly side of the cosmos.

Technical Summary: spore – An Event-Level Sampling Pipeline for Multi-Telescope Neutrino Astronomy

Problem Statement
The field of high-energy neutrino astronomy is transitioning into a multi-telescope and multi-energy era, involving simultaneous operations of detectors like IceCube, KM3NeT, P-ONE, TRIDENT, and IceCube-Gen2, alongside proposed future arrays (TAMBO, TRINITY, GRAND, HERON). Current simulation tools fail to adequately support event-level analysis across these diverse instruments. Existing end-to-end Monte Carlo pipelines (e.g., Prometheus) are computationally expensive and require proprietary reconstruction chains. Conversely, tools operating on aggregate distributions (e.g., toise) are suitable for design studies but cannot generate individual reconstructed events required for modern likelihood analyses. There is a critical need for an open-source, IRF-agnostic tool that can generate realistic, event-level data from tabulated instrument response functions (IRFs) for both point and extended sources across multiple detectors.

Methodology
The paper presents spore, a pure-Python package (requiring Python ≥3.11) designed to bridge this gap. The software architecture is modular, separating coordinate handling, physics definitions, source modeling, detector response loading, and event sampling.

Instrument Response Function (IRF) Framework: spore ingests detector responses in a standardized, selection-agnostic HDF5 format. This format encapsulates three core components:
1. Effective Area ( $A_{eff}$ ): Defined as a function of neutrino energy and zenith angle, distinguishing between track-like (primarily $\nu_\mu$ ) and cascade-like events.
2. Point Spread Function (PSF): Stored as an inverse cumulative distribution function (CDF) of the angular offset $\psi$ given true energy, allowing for the sampling of reconstructed directions.
3. Energy Resolution: Stored as an inverse CDF of the log-ratio $\Delta = \ln(E_{reco}/E_{true})$ , enabling the sampling of reconstructed energies.
  Note: The framework also supports joint distributions (e.g., $P(E_{reco}, \psi, \sigma_\psi | E_{true}, \delta)$ ) directly if provided.
Source Models: The package supports two primary source geometries:
1. Point Sources: Defined by fixed sky coordinates and a differential flux model (analytic power-law or tabulated).
2. Extended Sources: Defined by spatial templates (e.g., Galactic plane, dark matter halos) with flux dependent on energy, declination, and right ascension.
  Flux models can be isotropic or spatially dependent, with support for tabulated spectra in 1D, 2D, or 3D.
Sampling Algorithms:
- Point Source Sampler: Computes the cumulative distribution of the event rate over energy and samples true energies via inverse CDF. Reconstructed directions and energies are smeared using the PSF and energy resolution.
- Extended Source Sampler: Utilizes a hierarchical inverse-CDF approach on a 3D grid ( $\sin\delta$ , $\alpha$ , $\ln E$ ). To handle the non-factorizable joint distribution caused by the zenith-dependent effective area, the sampler precomputes weight functions and factorizes the joint distribution into conditional probabilities. It employs adaptive energy binning to resolve steep spectral features near detection thresholds.
- Multi-Detector Support: Both samplers accept lists of detectors. Each detector independently Poisson-samples its event count based on its specific effective area and geographic location, with events tagged by detector ID for joint analyses.
- Modes: Sampling can be performed in "Poisson mode" (stochastic event counts) or "fixed-N mode." Effective areas can be evaluated at a single epoch (instantaneous) or averaged over a diurnal cycle (steady-state).

Key Contributions

Event-Level Generation from Tabulated IRFs: spore is the first publicly available tool to generate individual reconstructed events (direction, energy, time, morphology, local zenith/azimuth) directly from tabulated IRFs without requiring proprietary simulation chains.
IRF Agnosticism: The package decouples simulation machinery from specific detectors. Any neutrino telescope with available $A_{eff}$ , PSF, and energy resolution can be integrated via an HDF5 file.
Native Multi-Telescope Support: The framework natively handles joint analyses of heterogeneous detector lists, allowing simultaneous sampling for instruments with distinct resolutions and locations.
Hierarchical Sampling: The extended source sampler avoids MCMC burn-in by using a precomputed hierarchical inverse-CDF grid, efficiently handling complex spatial flux maps.

Validation Results
The authors validated spore using two public IceCube data releases:

HESE 7.5-Year Release: The sampled deposited-energy spectrum for High-Energy Starting Events (HESE) tracks the published best-fit expectation. A comparison of 1,000 pseudo-experiments against the observed data shows the data falls within the goodness-of-fit distribution, confirming the sampler reproduces both the mean expectation and Poisson fluctuations.
IceCube 10-Year Tracks (Northern Sky):
- Declination Distribution: The sampled declination distribution for northern-sky tracks agrees with observed data to within 10–15%.
- Reconstructed Energy: The energy distribution agrees with data over the bulk of the sample (peak near 800 GeV up to tens of TeV). However, below 600 GeV, the sampler over-predicts the rate by up to a factor of three. The authors attribute this discrepancy to the coarse binning (0.5 in $\log_{10} E$ ) and declination averaging of the public smearing matrix used in the IRF, rather than a flaw in the sampling algorithm. This was confirmed by an independent forward-fold calculation yielding similar results.
Point Source Analysis ( $\psi^2$ ): A synthetic event sample including a signal at NGC-1068 was generated. The resulting $\psi^2$ distribution correctly reproduces the expected background shape (via RA scrambling) and signal excess, validating the PSF sampling and declination-dependent weighting.

Significance and Claims
The paper positions spore as a necessary tool for the evolving landscape of neutrino astronomy. Its primary significance lies in enabling event-level analysis for multi-detector studies without the computational overhead of full Monte Carlo simulations or the need for proprietary reconstruction software.

The authors modestly claim that spore fills a specific gap: no other public tool simultaneously (i) produces final-level reconstructed events, (ii) supports a broad range of source hypotheses within a single framework, and (iii) allows extension to new observatories simply by supplying a response file rather than modifying code. The paper explicitly notes limitations, such as the inability to simulate atmospheric muons (no IRF provided) and the dependence on the fidelity of input IRFs, particularly regarding the coarse binning of public smearing matrices which limits low-energy accuracy. Future work is outlined to expand the library of IRF files, support time-dependent effective areas, and provide higher-level sensitivity-scan interfaces.

More like this