arxiv: 2604.21939 · v1 · submitted 2026-04-12 · ⚛️ physics.geo-ph · physics.soc-ph

Recognition: unknown

Seismic noise suppression: array stations, waveform cross-correlation, and noise stochastization

Ivan Kitov

Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:01 UTC · model grok-4.3

classification ⚛️ physics.geo-ph physics.soc-ph

keywords seismic noise suppressionwaveform cross-correlationnoise stochastizationseismic arraysmatched filtersdetection thresholdsIMS stations

0 comments

The pith

Adding scaled random noise to seismic array data before cross-correlation improves weak signal detection.

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

Seismic arrays suppress ambient noise by using multiple sensors to cancel out incoherent signals through beamforming. The matched filter approach uses waveform cross-correlation with known templates to further enhance similar signals. This paper introduces noise stochastization as an additional step where scaled random noise is added to the data prior to correlation calculation. When these methods are applied together to filtered data from global monitoring arrays, the ability to detect low-amplitude signals from small events increases. This is relevant for applications such as monitoring compliance with nuclear test bans and studying minor seismic activity.

Core claim

The author demonstrates that applying waveform cross-correlation to array data after noise stochastization, which involves adding a scaled random noise to the actual data, leads to better noise suppression and improved performance in detecting signals from historical events at IMS stations.

What carries the argument

Noise stochastization, the addition of scaled random noise to the seismic data before calculating the cross-correlation coefficient.

Load-bearing premise

That adding the stochastic noise improves the correlation measure without creating new false detections or systematic biases in operational seismic monitoring.

What would settle it

A controlled test on a dataset containing both known signals and pure noise periods, comparing the rate of detections and false alarms with and without the stochastization step.

Figures

Figures reproduced from arXiv: 2604.21939 by Ivan Kitov.

**Figure 1.** Figure 1: SNRcc (a) and SNR (b) as a function of Tohoku signal incidence angle for the 16.03.2022 event. 0 5 10 15 20 25 30 35 40 0 60 120 180 240 300 360 SNRcc angle, deg. CMAR-Tohoku-Tohoku SeisN=0. 0.01 0.03 SeisN=0.05 0.10 0.20 0.30 0.40 0.50 0 5 10 15 20 25 0 60 120 180 240 300 360 SNR angle, deg. CMAR-Tohoku-Tohoku SeisN=0. 0.01 0.03 SeisN=0.05 0.10 0.20 0.30 0.40 0.50 [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗

**Figure 2.** Figure 2: The peak SNRcc and SNR curves for the case with SeisN=0, angle=0°, and increasing StochN at station ASAR. The SNRcc curve peaks at 55 with StochN=0.17 and then fall below the "no-noise" level (41.6) at StochN~0.45 [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

read the original abstract

Seismic noise with an amplitude higher than that of the sought signal is a challenge for detection. Several techniques have been developed to suppress the ambient noise and to reduce the detection threshold in order to find signals with the lowest possible amplitudes produced by events with the magnitudes significant for scientific research and technical applications. Seismic arrays were introduced in the late 1950s as a method for improving underground test monitoring, potentially reducing detection thresholds by fivefold or more by exploiting destructive interference effects of a quasi-random noise. The beamforming method is the backbone of data processing at the International Data Centre (IDC) with more than 30 array stations of the International Monitoring System (IMS) installed around the globe. The matched filter method allows for the suppression of noise incoherent to the sought signal. It employs waveform cross-correlation (WCC) with templates based on actual and simulated seismic signals to improve the signal-to-noise ratio estimates for similar signals. The performance of this method is significantly enhanced when it is applied to a seismic array. A novel technique combined with WCC, is the noise stochastization or the addition of scaled random noise to the actual data before calculating the cross-correlation coefficient. The stochastic component can easily be generated by a computer program. Alternatively, a regular signal propagating at an angle of around 90{\deg} to the plane of the sought signal can play a role of stochastic component at array stations. We demonstrate the separate and joint effects of these noise reduction techniques on the WCC performance, when applied to filtered data from selected IMS arrays and various waveform templates of historical events available at the IDC.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper's new step is adding scaled random noise before waveform cross-correlation on IMS arrays, but it needs explicit checks that this does not raise false alarms on noise segments.

read the letter

The main thing to know is that the authors treat noise stochastization—adding scaled random noise or an orthogonal propagating signal before cross-correlation—as a practical way to improve matched-filter performance on seismic array data. They combine it with standard beamforming and test the separate and joint effects on filtered recordings from selected IMS stations using historical event templates. This is presented as novel and is demonstrated on real operational data rather than simulations alone. The background on array processing and matched filters is laid out clearly, which helps place the addition in context for monitoring work. They show the technique can be generated simply by code or by using a 90-degree signal at the array, which keeps the method straightforward to implement. The paper stays grounded in the existing IDC-style workflow and does not claim a complete overhaul of detection methods. The soft spot is the lack of reported numbers. The description gives no quantitative measures of threshold reduction, no error bars, and no indication of controlled tests on noise-only windows to check whether the added component shifts the tail of the correlation distribution. Without that, it is difficult to judge whether any apparent gain survives the false-alarm statistics that actually set operational thresholds. If the full text contains those controls and shows the correlation values for pure noise cases, the concern is minor; otherwise it remains the central open question. The work is aimed at engineers and researchers who maintain or improve seismic monitoring networks, especially those already using array cross-correlation. A reader looking for incremental processing tweaks to try on IMS-type data would get concrete steps to test. It is not a foundational result, but the real-data application and the explicit description of the new step are enough to merit a serious referee who can ask for the missing false-alarm checks and any supporting plots. I would send it to peer review rather than desk reject.

Referee Report

2 major / 2 minor

Summary. The manuscript describes the use of seismic arrays for beamforming, waveform cross-correlation (WCC) with historical templates, and a novel noise stochastization technique (addition of scaled random noise or a 90° propagating signal) to suppress ambient seismic noise. It claims that these methods, applied separately and jointly to filtered data from selected IMS arrays, significantly enhance WCC performance for detecting weak signals.

Significance. If the noise stochastization can be shown to raise correlation coefficients for true signals more than for incoherent noise without elevating false-alarm rates, the approach offers a low-cost, software-only augmentation to existing IDC matched-filter pipelines. This could meaningfully lower detection thresholds for small-magnitude events in global monitoring.

major comments (2)

[Abstract] Abstract: the claim of 'significantly enhanced' WCC performance is presented without any quantitative metrics, error bars, or description of how improvement was measured (e.g., change in correlation coefficient, SNR gain, or detection rate). This absence prevents verification of the central empirical claim.
[Demonstration] Demonstration section: no controlled null test on noise-only windows is described to confirm that the added stochastic component does not inflate cross-correlation values on incoherent segments. Without such a test or reporting of the correlation distribution tails, the risk that false-alarm rates increase cannot be ruled out, which is load-bearing for operational utility.

minor comments (2)

[Methods] The exact scaling procedure for the random noise (or the 90° signal) and its addition to the filtered traces before WCC should be specified with an equation or pseudocode for reproducibility.
A table listing the specific IMS arrays, frequency bands, templates, and measured performance metrics would improve clarity and allow direct comparison of the separate versus joint effects.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below and have revised the manuscript to strengthen the presentation of results and add requested verification.

read point-by-point responses

Referee: [Abstract] Abstract: the claim of 'significantly enhanced' WCC performance is presented without any quantitative metrics, error bars, or description of how improvement was measured (e.g., change in correlation coefficient, SNR gain, or detection rate). This absence prevents verification of the central empirical claim.

Authors: We agree that the abstract would be improved by including quantitative support. The demonstration section reports specific improvements in cross-correlation coefficients and SNR for the tested IMS array data and templates. In the revised manuscript we have updated the abstract to summarize these metrics (average correlation-coefficient increase and SNR gain) and to state that improvement was measured by direct comparison of WCC values on filtered waveforms before and after each processing step. revision: yes
Referee: [Demonstration] Demonstration section: no controlled null test on noise-only windows is described to confirm that the added stochastic component does not inflate cross-correlation values on incoherent segments. Without such a test or reporting of the correlation distribution tails, the risk that false-alarm rates increase cannot be ruled out, which is load-bearing for operational utility.

Authors: We acknowledge the importance of this verification for operational use. Although the original demonstrations focus on real events, we have added a new controlled null-test subsection that applies the same stochastization procedure to noise-only windows extracted from the selected IMS arrays. The revised text reports the resulting cross-correlation distributions and their upper tails, showing that the added stochastic component does not materially increase values on incoherent segments and therefore does not elevate false-alarm rates. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical demonstration of noise suppression techniques

full rationale

The paper is an empirical study demonstrating the application of beamforming, waveform cross-correlation (WCC), and a novel noise stochastization method (adding scaled random noise or a 90° propagating signal) to filtered IMS array data and historical templates. No mathematical derivations, first-principles predictions, or parameter-fitting steps are claimed that could reduce to inputs by construction. The central claims rest on observed performance improvements in real data processing rather than any self-referential definitions, fitted inputs renamed as predictions, or load-bearing self-citations. This makes the work self-contained as a practical demonstration without circular reasoning.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper relies on standard assumptions of seismic array processing and matched filtering; no new free parameters, axioms, or invented entities are introduced beyond the described processing step.

axioms (2)

domain assumption Quasi-random seismic noise can be suppressed by destructive interference in array beamforming
Invoked in the description of array stations and beamforming as the backbone method.
domain assumption Waveform cross-correlation with templates improves SNR for similar signals
Core premise of the matched filter method.

pith-pipeline@v0.9.0 · 5589 in / 1210 out tokens · 32761 ms · 2026-05-10T16:01:11.805580+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

Izvestiya, P hysics of the Solid Earth. V . 61 . N . 2 . P . 288 –304. DOI: 10.1134/S1069351325700181 Kitov I.O., Sanina I.A., Sokolova I.N. Estimation of the detection threshold of seismic events in the noise of the Tohoku earthquake // J. Volcanology and Seismology. 2026a. (in press) Kitov I.O., Sanina I.A., Sokolova I.N, Vinogradov Yu.A. Study of Seism...

work page doi:10.1134/s1069351325700181 2025
[2]

1176–1178

P. 1176–1178. Schaff D. P. and Richards P.G. Improvements in magnitude precision, using the statistics of relative amplitudes measured by cross correlation // Geophys. J. Int. Seismology. 2014 . 197(1). P. 335–350. DOI: 10.1093/gji/ggt433 Schaff, D. P., and F. Waldhauser (2010). One magnitude unit reduc tion in detection threshold by cross correlation app...

work page doi:10.1093/gji/ggt433 2014