{"id":"46883508-ff41-4942-a4c6-ad340e5a4209","arxiv_id":"2507.13523","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Distributed acoustic sensing is shown to cancel seismometer noise with a residual factor near 0.11 at 20 Hz, comparable to geophone arrays, though the evaluation appears to use in-sample Wiener filtering.","lead":"This paper tests whether fiber-optic distributed acoustic sensing can monitor ground motion well enough to cancel Newtonian noise in gravitational wave detectors. It reports that DAS matches geophones in a noise-cancellation case study, but the evaluation has methodological gaps.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"In-sample Wiener filter estimation likely inflates reported 0.11 residual; no train/test split is described, and DAS vertical residual at 20 Hz is internally inconsistent (0.11 vs 0.14).","rationale":"The most load-bearing claim is the quantitative statement that DAS achieves a residual noise factor of 0.11 at 20 Hz, comparable to geophones. For this claim to hold, the Wiener filter must generalize to independent data. The manuscript never describes splitting the data into estimation and evaluation sets. The cross-spectral matrices in Eq. (13) are computed with Daniell's method on the same 1-hour record, and the residual is then derived from those same matrices. This is a textbook in-sample evaluation, which is known to be optimistically biased for multichannel Wiener filters because the filter can adapt to the specific noise realization. The bias grows with the number of channels (up to six in the main comparison) and with the number of frequency bins, making the 0.11 figure particularly suspect. The same weakness affects the geophone results, so the relative comparison may still be informative, but the absolute residual values cannot be trusted without out-of-sample validation. I agree with the reader's weakest assumption. An independent internal inconsistency—0.11 in Table 1 vs 0.14 in Section 5.2.1 for the same DAS vertical 20 Hz case—further weakens confidence in the reported numbers. Because the reader already issued a CONDITIONAL verdict conditioned on exactly this issue, my stress-test does not change the verdict; it strengthens the condition that out-of-sample validation is required. The concrete test described above would settle the concern: if the out-of-sample residual remains near 0.11, the DAS cancellation claim is credible; if it degrades substantially, the headline should be revised or the claims downgraded.","tokens_in":13304,"tokens_out":3766,"duration_ms":40564,"concrete_test":"Split the 1-hour record into two disjoint 30-minute segments. Estimate CDS, CDD, and CSS from the first segment, compute the Wiener filter, and evaluate Eq. (13) on the held-out second segment. Repeat with swapped segments and average the residuals. If the out-of-sample DAS residual at 20 Hz with six channels is substantially larger than the in-sample 0.11 (e.g., above 0.3), the reported cancellation factor is an artifact of in-sample fitting. Also reconcile the vertical DAS 20 Hz residual stated as 0.11 in Table 1 and 0.14 in Section 5.2.1 by re-examining the exact computation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.3 Eq. (13) defines the residual factor √R(ω) using cross-spectral matrices CDS, CDD, and CSS, estimated from the same 1-hour dataset on which residuals are reported; no train/test split or cross-validation is described in Sections 4.2, 4.3, or 5.2. The Wiener filter coefficients are therefore fitted to the same noise realization used to evaluate cancellation performance. With up to six DAS channels and many frequency bins, this in-sample procedure can substantially underestimate the achievable residual. The same issue affects the geophone comparison and the geophone-prediction correlation of 0.97 in Section 5.3, so the headline 'residual noise factor of 0.11' is not yet established as an out-of-sample cancellation factor. Additionally, the abstract and Table 1 give a DAS vertical residual of 0.11 at 20 Hz with six channels, while Section 5.2.1 states 0.14 for the same configuration, indicating an unresolved internal inconsistency.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that distributed acoustic sensing (DAS) is a viable and scalable alternative to seismometers and geophones for environmental monitoring and Newtonian-noise mitigation in gravitational-wave detectors. Using field data from the DESY campus (vibrotruck, earthquake, thunderstorm, HVAC), the authors convert DAS strain-rate measurements to ground acceleration using a semblance-estimated slowness, compare waveforms and power spectra with co-located broadband seismometers, use multichannel Wiener filtering to cancel the vertical and east components of a seismometer with DAS and with geophones, and report residual noise factors down to 0.11 at 20 Hz. They also report a 0.97 correlation for Wiener-filter prediction of geophone signals from DAS, DAS coherence lengths of roughly 11 m and 23 m in the 3-10 Hz band, and several environmental monitoring examples.","tokens_in":13575,"tokens_out":4224,"duration_ms":44661,"significance":"If the central cancellation claim is established out of sample, the paper would provide a practically important result: DAS could serve as a scalable witness sensor for Newtonian-noise cancellation in current and next-generation gravitational-wave detectors, using existing fiber infrastructure and dense spatial sampling. The paper also contributes useful field demonstrations of DAS-to-seismometer conversion, coherence-length measurements, and environmental monitoring. However, the headline cancellation performance is currently supported only by in-sample Wiener-filter residuals, and the paper contains an internal inconsistency in the reported DAS vertical residual (0.11 vs 0.14). Those issues directly affect the central claim, so the significance of the result is not yet established.","major_comments":[{"comment":"The residual factors reported in Table 1 and the abstract are computed from cross-spectral matrices CDS, CDD, and CSS estimated with Daniell's method on the same 1-hour dataset used for evaluation; no train/test split or cross-validation is described. Since the Wiener filter coefficients are fitted to the same noise realization that is being cancelled, the in-sample residual sqrt(R(ω)) is optimistically biased, and the headline 'residual noise factor of 0.11 at 20 Hz' is not established as an out-of-sample cancellation factor. Please report out-of-sample residuals (or k-fold cross-validation) with uncertainties, and specify the number of frequency bins/taps used in the Wiener filter.","section":"Section 4.3, Eq. (13); Section 5.2"},{"comment":"The abstract and Table 1 report a DAS vertical-component residual of 0.11 at 20 Hz with six channels, while Section 5.2.1 states that 'DAS achieved 0.14' for the same configuration; Section 5.2 also gives 0.14 for six DAS channels at 20 Hz. This internal inconsistency affects the central comparison and must be resolved before the headline claim can be accepted.","section":"Abstract; Table 1; Section 5.2.1"},{"comment":"The geophone-prediction correlation of 0.97 and the coherence 'above 0.7' are computed with Wiener filter coefficients estimated from the same data used for the prediction, so this metric is a fit-quality measure rather than a predictive validation. Please provide an out-of-sample evaluation and specify the Wiener filter length or number of taps, the frequency smoothing, and the dataset split.","section":"Section 4.2; Section 5.3"},{"comment":"The DAS-to-acceleration conversion relies on an apparent slowness estimated from the DAS data itself via semblance (Eq. 8), and the selected DAS channels used for cancellation are not specified in Section 4.3 or Table 1. Because both the slowness and the channel selection are adapted to the record being analyzed, they can inflate the reported correlation and cancellation performance; please describe the channel-selection criteria, report sensitivity to slowness and channel choice, and ideally validate the conversion with an independent slowness estimate from the geophone array.","section":"Section 4.1; Section 3.1.1; Section 6"}],"minor_comments":[{"comment":"The Gaussian correlation length Lc is a fitted parameter, and the reported coherence lengths (11 m and 23 m) come from a single vibrotruck event; please report the fit uncertainty and state whether the result is stable across events and frequency bands.","section":"Section 6.0.1"},{"comment":"The wavenumber vector is written as k = 2π s(f)·(sinθ, cosθ, 0), which appears dimensionally inconsistent with the slowness definition in Eq. (6); please define f and s explicitly and check the units.","section":"Equation (11)"},{"comment":"Reference [20] is cited for LPSD, but the citation appears to point to a software package rather than the original LPSD algorithm; please cite the original algorithm and provide a version/commit for the software if used.","section":"References"},{"comment":"There are several typographical and nomenclature issues, including 'Fracensca' in the author list, 'Deutches' in the acknowledgments, 'compliment' for 'complement' in Section 6, and the irregular spacing in 'W A VE'; these should be corrected.","section":"Throughout"},{"comment":"The PDF plot of residuals should specify how the residuals are aggregated over frequency and channels; the statement that the maximum lies between 0.00 and 0.02 needs error bars or sample counts to be meaningful.","section":"Figure 13"}],"recommendation":"major_revision","confidential_remarks":"The central claim is promising but currently rests on in-sample evaluation and an unresolved internal inconsistency. The authors should be encouraged to revise with out-of-sample validation and to clarify the channel-selection and slowness-estimation procedures before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe new element here is the first reported case of DAS actually cancelling a colocated seismometer's noise, with residual factors comparable to a geophone array, plus coherence-length measurements from a 19-km fiber at DESY. That is a genuine experimental step, not just another equivalence study. The paper also does some things well: the vibrotruck and earthquake events give real waveforms, the PSD and Bland–Altman checks are reasonable for sensor comparison, and the coherence-length numbers (11 m and 23 m in the two bands) are concrete and match Gaussian fits.\n\nThe problem is the evaluation of the cancellation residual. The Wiener filter coefficients in Eq. (13) are estimated from the same 1-hour dataset used to report the residual. There is no train/test split, no cross-validation, nothing in Sections 4.2, 4.3, or 5.2 that says the filter was applied to a held-out segment. So the headline “residual noise factor 0.11” at 20 Hz is an in-sample number, and with up to six channels and many frequency bins it will be optimistically biased. The geophone comparison is subject to the same issue, so the claim of comparability is not yet established out of sample.\n\nThere is also an internal inconsistency: the abstract and Table 1 give 0.11 for the DAS vertical residual at 20 Hz with six channels, while Section 5.2.1 states 0.14 for the same configuration. That needs reconciliation. Less serious, but still relevant: the DAS channel selection is not specified (which channels, how many, how chosen), and the slowness used for DAS-to-acceleration conversion is itself fitted from the DAS data, adding another free parameter. No error bars or statistical significance are given for the residuals.\n\nThe central concept is sound and the experiment is real. The in-sample estimate is the main soft spot, not a fatal flaw—it can be fixed by re-estimating the filter on a training segment and evaluating on a test segment. The coherence-length analysis and the environmental monitoring sections are more straightforward and hold up.\n\nWho is this for? People working on Newtonian noise mitigation for Einstein Telescope or LIGO upgrades, and DAS applications in seismology. It deserves a serious referee. I'd send it to peer review with a clear request for out-of-sample validation and a resolution of the 0.11/0.14 discrepancy. The paper's usefulness depends on whether the residuals survive out-of-sample testing.","headline":"First DAS-based cancellation of a colocated seismometer's noise is a real experimental step, but the headline residual is in-sample and internally inconsistent.","tokens_in":14062,"tokens_out":2727,"would_cite":false,"duration_ms":27506,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Distributed acoustic sensing, using ordinary optical fiber, cancels a co-located seismometer's noise to the same residual factor as a geophone array (0.11 at 20 Hz), making it a scalable witness-sensor option for gravitational wave…","keywords":["distributed acoustic sensing","Newtonian noise mitigation","gravitational wave detectors","Wiener filter","seismometer comparison","geophone array","coherence length","environmental monitoring"],"falsifier":"Train the multichannel Wiener filter on the first half of the one-hour dataset and apply it to the second half, then recompute the residual factor at 20 Hz for the vertical seismometer component using six DAS channels; a held-out residual substantially above 0.11 (for example, above 0.3) would show the headline cancellation factor does not generalize.","tokens_in":13171,"feed_emoji":"🔭","tokens_out":11015,"duration_ms":111162,"temperature":0.7,"pith_summary":"Distributed acoustic sensing (DAS) turns a standard optical fiber into thousands of closely spaced strain sensors, and this paper argues that those sensors can match conventional seismometers and geophones for cancelling Newtonian noise—the gravity-gradient disturbance that limits the low-frequency sensitivity of gravitational wave detectors. Dense seismometer arrays are expensive and spatially sparse, while fiber is already installed at many observatory sites, so a comparable DAS performance would make witness-sensor networks far easier to scale. The evidence is a direct comparison on a research campus: DAS and colocated seismometer recordings correlate above 0.8 in the 3–20 Hz band, DAS-predicted geophone signals keep coherence above 0.7, and a multichannel Wiener filter fed by six DAS channels reduces a colocated seismometer's vertical-component noise to a residual factor of 0.11 at 20 Hz, equal to the geophone array. If the result holds, DAS becomes a practical complement to traditional seismic instruments for both environmental monitoring and low-frequency noise suppression in current and next-generation detectors.","feed_headline":"DAS matches geophone arrays for Newtonian-noise cancellation","feed_subtitle":"Strain-sensing fiber matches geophones: 0.11 residual noise at 20 Hz, a scalable route to quieter detectors.","key_machinery":"The mechanism is the equivalence between DAS strain-rate and ground velocity: $d\\varepsilon/dt = [v(x+L_g/2)-v(x-L_g/2)]/L_g$, with the conversion $\\varepsilon = du/dx = \\pm (1/s)\\,du/dt$, where $s$ is the local apparent slowness estimated by semblance analysis, a coherence-based scan over trial slowness values. Dividing the DAS strain-rate by $s$ yields ground acceleration comparable to a seismometer. On this sits the multichannel Wiener filter, whose residual is $R(\\omega)=1-\\vec{C}_{DS}^{\\dagger}(\\omega)\\vec{C}_{DD}^{-1}(\\omega)\\vec{C}_{DS}(\\omega)/C_{SS}(\\omega)$; the square root $\\sqrt{R}$ is the reported noise-reduction factor. The gauge length $L_g$ acts as a spatial averaging scale, and the dense channel spacing lets the array reconstruct wavelengths longer than a single gauge length.","core_discovery":"On the paper's own terms, the central discovery is that DAS is not just a detector of seismic events but a quantitatively equivalent seismic instrument. After converting DAS strain-rate to ground acceleration through the apparent slowness, the DAS waveforms, power spectra, and Bland–Altman agreement match a colocated broadband seismometer. Used as witness channels in a multichannel Wiener filter, six DAS sensors cancel a seismometer's vertical noise to a residual factor of 0.11 at 20 Hz, exactly matching six geophone channels, and outperform the geophones on the horizontal component (0.04 versus 0.15). The paper also reports that this high colocated correlation contradicts a prior theoretical expectation that strain meters must be placed far from the test mass to be useful for Newtonian-noise cancellation.","pith_inferences":["A held-out evaluation is the natural next step: the paper's residual factors are in-sample, so a train/test split would give a realistic bound on DAS cancellation performance.","Because DAS measures strain only along the fiber direction, the vertical-component improvement with more channels suggests that helical or multi-orientation cable layouts could push vertical cancellation below the quoted 0.11; that is a direct, testable extension.","The reported coherence lengths of roughly 11 m and 23 m in the 3–10 Hz band imply an optimal witness-channel spacing for cancellation; arrays much denser than that may add little information while raising processing cost.","If the colocated high correlation survives at quieter detector sites, existing telecommunication fiber around gravitational wave observatories could be repurposed for Newtonian-noise monitoring, with the open engineering question being real-time slowness estimation and adaptive filter updates."],"forward_implications":["Existing fiber infrastructure can be turned into dense seismic arrays, reducing the deployment cost of witness sensors for Newtonian-noise cancellation at gravitational wave observatories.","DAS matches a geophone array for vertical-component cancellation (0.11 residual at 20 Hz with six channels) and exceeds it on the horizontal component (0.04 versus 0.15).","DAS can reconstruct geophone signals with an average correlation of 0.97 and coherence above 0.7 above 3 Hz, allowing it to substitute for or supplement sparse geophone networks.","DAS coherence lengths of about 11 m at 3–5 Hz and 23 m at 6–10 Hz resolve ground-motion structure finer than typical 25 m geophone spacing, which is relevant near detector test masses.","Combining DAS with traditional seismometers and geophones is identified as a promising direction for further improving Newtonian-noise cancellation and environmental monitoring."],"supporting_citations":[{"why":"Supplies the operating principle of DAS—Rayleigh backscattering, gauge length, and strain measurement—on which the conversion to seismometer-equivalent units is built.","marker":"[1]"},{"why":"Defines terrestrial gravity fluctuations and Newtonian noise, and states the far-field expectation about strain meters that the paper's colocated-correlation result directly challenges.","marker":"[2]"},{"why":"Establishes seismic gravity-gradient noise as a fundamental low-frequency limit for interferometric gravitational-wave detectors, motivating the cancellation study.","marker":"[5]"},{"why":"Demonstrates co-located DAS and seismometer array recordings and gives the strain-to-ground-motion relationship used to convert DAS output to acceleration.","marker":"[9]"},{"why":"Provides simultaneous DAS and geophone field measurements that ground the DAS-to-geophone signal relationship used in the Wiener-filter reconstruction.","marker":"[10]"},{"why":"Characterizes the broadband instrument response of fiber-optic DAS arrays, supporting treatment of DAS output as a calibrated strain-rate measurement.","marker":"[11]"},{"why":"Gives the complex semblance formula the paper implements to estimate apparent slowness from the DAS array.","marker":"[15]"},{"why":"Documents geophone-based Newtonian-noise cancellation at operating gravitational wave detectors and supplies the multichannel Wiener-filter formulation and baseline residuals.","marker":"[19]"}],"fun_headline_variants":["DAS matches seismometers for Newtonian noise","Fiber sensing rivals geophones for detector noise","DAS cancels seismic noise: 0.11 residual at 20 Hz","DAS: scalable seismic monitoring for gravitational waves","Distributed sensing matches geophones in noise cancellation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The cancellation filter is tuned and evaluated on the same one-hour dataset, with no separate test period mentioned, so the reported residual factors may be optimistically biased by in-sample fitting.","fun_headline_variants_meta":{"raw":{"variants":["DAS matches seismometers for Newtonian noise","Fiber sensing rivals geophones for detector noise","DAS cancels seismic noise: 0.11 residual at 20 Hz","DAS: scalable seismic monitoring for gravitational waves","Distributed sensing matches geophones in noise cancellation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000124,"raw_usage":{"total_tokens":1111,"prompt_tokens":963,"completion_tokens":148,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":579,"completion_tokens_details":{"reasoning_tokens":70}},"tokens_in":579,"tokens_out":148,"duration_ms":2424,"temperature":1.0,"reasoning_tokens":70,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:23:00.140084+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Train the multichannel Wiener filter on the first half of the one-hour dataset and apply it to the second half, then recompute the residual factor at 20 Hz for the vertical seismometer component using six DAS channels; a held-out residual substantially above 0.11 (for example, above 0.3) would show the headline cancellation factor does not generalize.","supporting_citations":[{"cited_title":"An Introduction to Distributed Optical Fibre Sensors","cited_arxiv_id":null,"evidence_quote":"Supplies the operating principle of DAS—Rayleigh backscattering, gauge length, and strain measurement—on which the conversion to seismometer-equivalent units is built."},{"cited_title":"Terrestrial gravity fluctuations","cited_arxiv_id":null,"evidence_quote":"Defines terrestrial gravity fluctuations and Newtonian noise, and states the far-field expectation about strain meters that the paper's colocated-correlation result directly challenges."},{"cited_title":"Seismic gravity-gradient noise in interferometric gravitational-wave detectors","cited_arxiv_id":null,"evidence_quote":"Establishes seismic gravity-gradient noise as a fundamental low-frequency limit for interferometric gravitational-wave detectors, motivating the cancellation study."},{"cited_title":"Ground motion response to an ML 4.3 earthquake using co-located distributed acoustic sensing and seismometer arrays","cited_arxiv_id":null,"evidence_quote":"Demonstrates co-located DAS and seismometer array recordings and gives the strain-to-ground-motion relationship used to convert DAS output to acceleration."},{"cited_title":"Field testing of modular borehole monitoring with simultaneous distributed acoustic sensing and geophone vertical seismic profiles at Citronelle","cited_arxiv_id":null,"evidence_quote":"Provides simultaneous DAS and geophone field measurements that ground the DAS-to-geophone signal relationship used in the Wiener-filter reconstruction."},{"cited_title":"On the Broadband Instrument Response of Fiber-Optic DAS","cited_arxiv_id":null,"evidence_quote":"Characterizes the broadband instrument response of fiber-optic DAS arrays, supporting treatment of DAS output as a calibrated strain-rate measurement."},{"cited_title":"Complex Semblance and Its Application","cited_arxiv_id":null,"evidence_quote":"Gives the complex semblance formula the paper implements to estimate apparent slowness from the DAS array."},{"cited_title":"Newtonian Noise Studies in 2nd and 3rd Generation Gravitational Wave In- terferometric Detectors","cited_arxiv_id":null,"evidence_quote":"Documents geophone-based Newtonian-noise cancellation at operating gravitational wave detectors and supplies the multichannel Wiener-filter formulation and baseline residuals."}],"review_version":1}