New Public Neutrino Alerts for Clusters of IceCube Events

Elisa Bernardini (for the IceCube Collaboration); Sarah Mancina; Sergio Cuenca

REVIEW 3 major objections 5 minor 13 references

New Public Neutrino Alerts for Clusters of IceCube Events

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The IceCube Collaboration is making its realtime cluster neutrino alerts public, replacing the private single-threshold stream with a three-tier system that reveals a source's full above-threshold behavior.

desk verdict A clearly written operations note on making GFU-cluster alerts public; the multi-threshold scheme is a design update, but the FAR curve needs error bars before the high-threshold rate is credible. read the letter →

arxiv 2507.07491 v1 pith:JXQ4T2RO submitted 2025-07-10 astro-ph.HE astro-ph.IM

Sarah Mancina , Sergio Cuenca , Elisa Bernardini (for the IceCube Collaboration) This is my paper

classification astro-ph.HEastro-ph.IM

keywords IceCubeneutrinoalertsGamma-rayFollow-Upclustermulti-messengerastrophysicsfalsealarmratereal-timeGCN

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper announces that IceCube's Gamma-ray Follow-Up (GFU) cluster alerts, which flag clusters of muon-neutrino events that could be flaring astrophysical sources, will move from private distribution to public distribution over NASA GCN and a dedicated website. It proposes a three-threshold alert scheme—low (3\$\sigma$), medium (3.75\$\sigma$), and high (4.5\$\sigma$)—calibrated to false alarm rates of roughly one per year and one per 20 years for the upper two thresholds. The key behavioral change is that the muting scheme, which previously suppressed updates from a source after the first alert, will be removed so that the full above-threshold activity of a monitored source remains visible. The central claim is that these changes make the cluster alert stream a more complete and open realtime multi-messenger resource without losing the statistical calibration that made the private alert stream usable.

What carries the argument

The load-bearing object is the time-dependent unbinned maximum likelihood ratio (Eq. 2) that fits the number of signal events, spectral index, and flare start time for each cluster, giving a test statistic converted to a pre-trial significance. On top of that, the new alert scheme is governed by a false-alarm-rate calibration (Fig. 3) that maps pre-trial significance thresholds (3\$\sigma$, 3.75\$\sigma$, 4.5\$\sigma$) to expected background alert rates, and dissemination runs through NASA GCN notices plus a Flask-based website that plots evolving significance curves.

What would settle it

Count the actual number of low, medium, and high alerts issued after the public stream launches and compare with the predicted background rates of about 20 per year, 1 per year, and 1 per 20 years for catalog sources not associated with real astrophysical activity; a statistically significant excess of medium or high alerts over several years would show the scrambled-time calibration underestimates the false-alarm rate. A more direct check is to rerun the background simulation with re-simulated detector livetime and seasonal variations rather than time scrambling and compare the test-statistic distributions.

Watch

Extended reading notes

Core claim

The GFU-cluster alert system identifies neutrino flares by applying a time-dependent unbinned maximum likelihood fit to events passing a signal-to-background event weight cut, computing a pre-trial significance for each incoming analysis-triggering event. The paper's central proposal is to publish these alerts publicly, replacing the single 3\$\sigma$ private alert threshold with a low/medium/high threshold ladder and streaming all above-threshold cluster information in realtime. This removes the previous muting behavior and lets external observers track how a source's significance develops over the 180-day window after a flare begins.

Load-bearing premise

The whole threshold ladder rests on the assumption that background simulations made by scrambling event times reproduce the true null distribution of the test statistic, and that the 11.5-year simulation sample represents future detector conditions; if scrambled times retain spurious temporal correlations or miss seasonal live-time variations, the quoted false-alarm rates (about one per year at 3.75\$\sigma$ and one per 20 years at 4.5\$\sigma$) will not hold.

Editorial extensions

If this is right

Every cluster crossing 3\sigma becomes public immediately, so the community sees a source enter and leave its active state rather than only the first crossing.
The medium and high tiers give a graded urgency signal, so follow-up telescopes can prioritize rare high-significance flares over routine 3\sigma crossings.
Removing the muting scheme means offline analyses using the public data no longer have a blind period that could bias source-catalog studies.
The same three-threshold structure can be applied to the all-sky mode once its higher trial factor is calibrated, extending public cluster alerts to unmonitored regions.
Joint cluster alerts with other neutrino telescopes become feasible because the realtime stream is openly available and not restricted by memoranda of understanding.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the source-list thresholds perform as calibrated, the all-sky mode will need much higher thresholds to reach comparable false-alarm rates, and the website design (significance curves) will likely become the standard visualization for time-dependent neutrino alerts across the field.
The removal of muting may reduce statistical blindness in offline catalog searches, but it also means the public stream itself becomes part of the trial factor for any retrospective search, a tension the paper does not address.
A natural testable extension is to use the medium and high alert rate over the first two years of public operation as an end-to-end check of the false-alarm computation, independent of background-only simulations.
One could also probe whether the Seyfert-type sources that motivate the public shift actually appear more often in cluster alerts than in single-event Gold and Bronze streams.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is a solid little operations paper, not a physics result. The genuinely new content is the plan to replace the private IACT-only GFU-cluster alerts with public GCN notices and a live website, using three significance thresholds (3, 3.75, 4.5 sigma) and unmuting the source behavior after the first alert. The statistical engine is the established unbinned likelihood from Braun et al. and Kintscher's thesis, so there is no new method. The paper is honest that the system is not yet operational and that fig 3 is simulation-only.

The best parts: the writing is clear, the design choices are explained, and the simulated examples in figs 1-2 illustrate how the significance curve behaves. They also correctly point to a companion paper for the source list design.

Soft spots, in proportion: the load-bearing FAR curve in fig 3 has no error bars and no statement of the number of simulated realizations. The claim that the 4.5 sigma threshold produces one false alert per 20 years cannot be supported by a one-year simulation alone; even with many realizations, the Poisson uncertainty at 0.05/yr is large. The low threshold (3 sigma) is anchored by actual operating experience since 2019 (~20/yr), so that part is credible, but the medium and high thresholds are extrapolations. The time-scrambling null also assumes the background is time-stationary; the paper should validate the FAR against real data, e.g., by checking that the observed 3-sigma rate matches the prediction.

One more minor point: the previous muting scheme existed to avoid biasing offline analyses, and the paper removes it without discussing how the collaboration will handle that blindness tradeoff. That might matter to IceCube's own later analyses.

Overall, I agree with the reader's conditional take. The paper is a useful status report for the MMA community, and it deserves a serious referee, but only if the referee asks for the missing calibration details. For a proceedings, this is probably fine; for a journal, I would want the FAR curve with confidence bands and a validation test before the advertised high-threshold rate is trusted.

Referee Report

3 major / 5 minor

Summary. This ICRC proceedings paper describes a plan to make IceCube's Gamma-ray Follow-Up (GFU) cluster alerts public, replacing the current private sharing with IACT partners and the muting scheme that suppresses updates after the initial alert. The proposed system uses three pre-trial significance thresholds (low 3σ, medium 3.75σ, high 4.5σ) chosen from the false alarm rate (FAR) curve derived from one year of background simulations (Figure 3). Alerts will be distributed via NASA GCN Notices and a dedicated website that provides continuous significance curves and alert histories. The paper focuses on the source-list mode, with all-sky mode deferred to a later release.

Significance. If the proposed system works as described, it would be a valuable community resource, enabling broader multi-messenger follow-up and removing the information-hiding muting scheme. The paper's main contribution is operational rather than scientific: it introduces a concrete multi-threshold public alert scheme and a real-time website. The design is clearly connected to previously validated likelihood methods, and the paper is transparent about the 'simulation only' status of the key calibration. However, the central calibration claim—that the chosen thresholds deliver the advertised false alarm rates—is not yet supported by the evidence presented, so the significance of the paper in its current form is that of a promising proposal rather than a validated system.

major comments (3)

[Section 2.3 and Figure 3] The FAR curve in Figure 3 is based on 'background simulations of 1 year of data' but shows no error bars and the text does not state the number of independent simulation realizations. For the high threshold (4.5σ), the claimed FAR is about 0.05 yr^-1, so a single one-year simulation would produce an expected 0.05 crossings; observing zero crossings cannot statistically distinguish this from a rate of order 1 yr^-1 (the 95% Poisson upper limit from zero events is about 3 yr^-1). The medium threshold (3.75σ, ~1 yr^-1) is also only weakly constrained. The authors should provide the simulation statistics, error bars, and a validation using the 2019–2024 operating data for the low threshold, and a concrete plan for validating the medium and high thresholds (e.g., via archival data or a larger Monte Carlo sample). Without this, the multi-threshold alert logic is not quantitatively established.
[Section 2.1 (null hypothesis construction)] The pre-trial p-values are computed by comparing the test statistic to distributions from background simulations 'generated by scrambling the data in time.' The paper does not demonstrate that this scrambling preserves the relevant null properties—such as detector livetime, seasonal modulation, and possible steady emission from catalog sources—or that the resulting TS distribution matches the true null. A mismatch would bias the FAR calibration in Figure 3 and therefore the alert thresholds. The authors should compare the simulated TS distribution with the distribution observed in archival GFU data over the same source catalog to validate the null construction.
[Sections 2.2 and 2.3] The paper proposes to remove the muting scheme to provide continuous updates on active sources, but it does not address how this is consistent with the stated original purpose of muting: preserving blindness for offline analyses that use similar source catalogs. The authors should either clarify the statistical policy for public updates versus internal offline searches, or explicitly state that the blindness concern is no longer considered relevant. This is not merely a presentation issue because it affects whether the public alert stream can coexist with IceCube's internal analysis program.

minor comments (5)

[Section 2.1] Equation (1): 'Raleigh distribution' should be 'Rayleigh distribution'.
[Figure 2 caption] The phrase 'all though they appearsimilar' contains a typo; it should be 'although they appear similar'.
[Section 2.2] The sentence 'this threshold was chosen to be 3σ and result in around 10 alerts per year being sent to each IACT (20 per year overall)' is grammatically awkward; also, clarify whether this rate is per source or summed over the source list.
[References] Reference [9] is incomplete: it lists 'IceCube Collaboration, R. Abbasi et al.' with no title, journal, or year.
[Section 3.2] The description of the website as 'built using the Flask framework' is an implementation detail; its relevance to the scientific content is unclear.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the false-alarm thresholds are a design calibration from background simulations, not a derived quantity recycled from fitted data.

full rationale

The paper's central claims are operational: choosing 3 sigma, 3.75 sigma, and 4.5 sigma thresholds to yield approximately 20, ~1, and ~1/20 background alerts per year respectively (Section 2.3, Figure 3). These are calibrated alert-stream design choices obtained from scrambled-time background simulations, not predictions recovered from fitted parameters. Pre-trial p-values are computed by comparing the likelihood test statistic to the null-hypothesis TS distribution from background simulations, which is independent of the threshold values; the false alarm rate is then simply the rate at which that same null distribution crosses a chosen threshold. Using the same simulated null for both p-values and FAR is the standard definition of a well-calibrated threshold, not a circular reduction. No fitted quantity from real data is renamed as a prediction, and no equation in the paper defines a target result in terms of itself. The self-citations present (e.g., [6] for the 2019 algorithm upgrade and [12] for the source-list design) describe the existing system and proposed catalog; the threshold scheme does not reduce to these citations, and no uniqueness theorem or ansatz is imported from them. The potential weakness that the one-year background simulation may not reliably support the claimed 1-in-20-year high-threshold rate is a validation and statistical-uncertainty concern about the calibration, not a circularity in the derivation chain. Accordingly, the analysis contains no self-definitional, fitted-input-as-prediction, or load-bearing self-citation step.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The plan rests on standard statistical assumptions plus two hand-chosen design parameters (alert thresholds and the fixed spectral index in the event weight). No new physical entities are introduced.

free parameters (2)

Low/medium/high alert thresholds = 3.0/3.75/4.5 sigma
Chosen by hand to match target false alarm rates of about 20, 1, and 0.05 per year from background simulations (Section 2.3).
Spectral index for event-weight signal PDF = -3.0
Fixed value used in Equation 1 to compute the initial event weight; later floated in the likelihood fit. This is an algorithmic design choice, not data-fitted.

assumptions (3)

domain assumption Time-scrambled data produce a valid null-hypothesis distribution for the test statistic
The pre-trial p-values and FAR curves in Figure 3 rely on scrambling event times in data to generate the background TS distribution (Section 2.2). If scrambling does not remove temporal correlations, p-values are biased.
domain assumption The signal PDF, a Rayleigh spatial term times a power-law energy term, describes neutrino flares from point sources
Equation 1 and the likelihood in Equation 2 assume this functional form; deviations would affect cluster significance.
domain assumption The GFU event selection and background PDF rates remain stable across the simulated 11.5-year period
The significance curve in Figure 2 and FAR estimates extrapolate from historical data to future operations.

how reviews work

0 comments

Cite this review

Pith. "Pith review of New Public Neutrino Alerts for Clusters of IceCube Events." pith.science (2026). https://pith.science/paper/JXQ4T2RO

@misc{pith2026250707491,
  author       = {Pith},
  title        = {Pith review of: New Public Neutrino Alerts for Clusters of IceCube Events},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JXQ4T2RO}},
  note         = {Machine review of arXiv:2507.07491}
}

read the original abstract

The IceCube Neutrino Observatory searches for the origins of astrophysical neutrinos using various techniques to overcome the significant backgrounds produced by cosmic-ray air showers. One such technique involves combining the neutrino data with other cosmic messengers to identify spatial and temporal correlations. IceCube contributes to multi-messenger astrophysics (MMA) by providing alerts for interesting events observed in the detector. The Gamma-ray Follow-Up (GFU) cluster alert system is one stream that identifies potential neutrino flares in realtime, producing around 20 alerts per year. GFU-cluster alerts have been privately shared with Imaging Air Cherenkov Telescopes (IACTs) through memoranda of understanding since IceCube's predecessor, AMANDA. To preserve blindness to the full behavior of our data, the current system mutes updates from sources following the initial GFU-cluster alert sent, preventing further updates until the activity drops below the alert threshold. With growing knowledge of the potential environments that produce astrophysical neutrinos and to foster open collaboration, the GFU-cluster alerts will shift to be publicly shared. Additionally, the new alert platform will provide all above-threshold information such that the source behavior after the initial alert is not obscured. The above threshold data will be distributed through an interactive website that will update the community on the status of active GFU-cluster alerts. This presentation will introduce the new GFU-cluster platform and the accompanying website, soon to be accessible to the MMA community.

Figures

Figures reproduced from arXiv: 2507.07491 by the authors.

**Figure 1.** Example of the incoming GFU event weights over 180 days (units in Modified Julian Days) around a source at declination -5.69°. The blue lines come from a background simulation derived from real data, with event times randomly shuffled. In orange are 5 signal events that were injected to mimic a flare from the source with a 50-day duration and a spectral index of -2.0. This 180-day period represents the data that the… view at source ↗

**Figure 2.** Example significance curve over 11.5 years of data for a simulated source at 𝛿 = 5.69°. The left triangles are the pre-trial significances for clusters from incoming analysis-triggering events plotted at their 𝑡1. The color represents the duration of the fitted time window in days (𝑡1 − 𝑡0). Clusters that would trigger alerts by crossing the significance thresholds are highlighted by the dashed circles with colors m… view at source ↗

**Figure 3.** False alarm rate (FAR) from background simulations of 1 year of data as a function of the alert threshold for single sources (dashed lines) and the sum of all sources in the full catalog proposed in [12]. The three thresholds proposed for the new alert scheme are highlighted by the dashed vertical lines and arrows. 2.3 Proposed Changes to Alert Scheme The new GFU-cluster platform will provide updates on the active s… view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Example source detail page from the GFU-cluster alerts webpage. The source location and details are shared at the top, and the significance curve (note: this is the same simulated source as shown in fig. 2). The significance curve figure is dynamic and can zoom into re…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

[1]

IceCubeCollaboration, M. G. Aartsenet al. JINST 12no. 03, (2017) P03012

work page 2017
[2]

IceCube, Fermi-LAT, MAGIC, AGILE, ASAS-SN, HA WC, H.E.S.S., INTEGRAL, Kanata, Kiso, Kapteyn, Liverpool Telescope, Subaru, Swift NuSTAR, VERITAS, VLA/17B-403 Collaboration, M. G. Aartsenet al. Science 361no. 6398, (2018) eaat1378

work page 2018
[3]

Abbasiet al

IceCubeCollaboration, R. Abbasiet al. Astrophys. J. Suppl. 269 no. 1, (2023) 25

work page 2023
[4]

Neutrino Triggered Target of Opportunity (NToO) test run with AMANDA-II and MAGIC

IceCube, MAGICCollaboration, M. Ackermann, E. Bernardini, N. Galante, F. Goebel, M. Hayashida, K. Satalecka, M. Tluczykont, and R. M. Wagner, “Neutrino Triggered Target of Opportunity (NToO) Test Run with AMANDA-II and MAGIC,” in30th International Cosmic Ray Conference (ICRC): Mérida, Mexico, vol. 3, pp. 1257–1260. 2007. arXiv:0709.2640 [astro-ph]

work page Pith review arXiv 2007
[5]

IceCubeCollaboration, M. G. Aartsenet al. Astropart. Phys. 92(2017) 30–41

work page 2017
[6]

Kintscher,Rapid Response to Extraordinary Events: Transient Neutrino Sources with the IceCube Experiment

T. Kintscher,Rapid Response to Extraordinary Events: Transient Neutrino Sources with the IceCube Experiment. PhD thesis, Humboldt-Universität zu Berlin, Mathematisch-Naturwissenschaftliche Fakultät, 2020

work page 2020
[7]

Kheirandish, K

A. Kheirandish, K. Murase, and S. S. KimuraAstrophys. J. 922 no. 1, (2021) 45

work page 2021
[8]

Abbasiet al

IceCubeCollaboration, R. Abbasiet al. Science 378 no. 6619, (2022) 538–543

work page 2022

Show all 13 references

[9]

Abbasiet al

IceCubeCollaboration, R. Abbasiet al
[10]

S. D. Barthelmy, P. Butterworth, T. L. Cline, N. Gehrels, G. J. Fishman, C. Kouveliotou, and C. A. MeeganAstrophysics and Space Science 231 no. 1-2, (Sept., 1995) 235–238

1995
[11]

Braun, J

J. Braun, J. Dumm, F. De Palma, C. Finley, A. Karle, and T. MontaruliAstropart. Phys. 29 (2008) 299–305

2008
[12]

Boscolo Meneguoloet al

IceCubeCollaboration, C. Boscolo Meneguoloet al. PoS ICRC2025(2025) (these proceedings) 919

2025
[13]

Grinberg,Flask web development

M. Grinberg,Flask web development. O’Reilly Media, 2 ed., Mar., 2018. 8 IceCube Public GFU-Cluster Alerts Full Author List: IceCube Collaboration R. Abbasi16, M. Ackermann63, J. Adams17, S. K. Agarwalla39, a, J. A. Aguilar10, M. Ahlers21, J.M. Alameddine22, S. Ali35, N. M. Ami...

2018

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

IceCubeCollaboration, M. G. Aartsenet al. JINST 12no. 03, (2017) P03012

work page 2017

[2] [2]

IceCube, Fermi-LAT, MAGIC, AGILE, ASAS-SN, HA WC, H.E.S.S., INTEGRAL, Kanata, Kiso, Kapteyn, Liverpool Telescope, Subaru, Swift NuSTAR, VERITAS, VLA/17B-403 Collaboration, M. G. Aartsenet al. Science 361no. 6398, (2018) eaat1378

work page 2018

[3] [3]

Abbasiet al

IceCubeCollaboration, R. Abbasiet al. Astrophys. J. Suppl. 269 no. 1, (2023) 25

work page 2023

[4] [4]

Neutrino Triggered Target of Opportunity (NToO) test run with AMANDA-II and MAGIC

IceCube, MAGICCollaboration, M. Ackermann, E. Bernardini, N. Galante, F. Goebel, M. Hayashida, K. Satalecka, M. Tluczykont, and R. M. Wagner, “Neutrino Triggered Target of Opportunity (NToO) Test Run with AMANDA-II and MAGIC,” in30th International Cosmic Ray Conference (ICRC): Mérida, Mexico, vol. 3, pp. 1257–1260. 2007. arXiv:0709.2640 [astro-ph]

work page Pith review arXiv 2007

[5] [5]

IceCubeCollaboration, M. G. Aartsenet al. Astropart. Phys. 92(2017) 30–41

work page 2017

[6] [6]

Kintscher,Rapid Response to Extraordinary Events: Transient Neutrino Sources with the IceCube Experiment

T. Kintscher,Rapid Response to Extraordinary Events: Transient Neutrino Sources with the IceCube Experiment. PhD thesis, Humboldt-Universität zu Berlin, Mathematisch-Naturwissenschaftliche Fakultät, 2020

work page 2020

[7] [7]

Kheirandish, K

A. Kheirandish, K. Murase, and S. S. KimuraAstrophys. J. 922 no. 1, (2021) 45

work page 2021

[8] [8]

Abbasiet al

IceCubeCollaboration, R. Abbasiet al. Science 378 no. 6619, (2022) 538–543

work page 2022