Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Daily Groundwater Monitoring Using Vehicle-DAS Elastic Full-waveform Inversion

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper claims that daily elastic full-waveform inversion of vehicle-generated surface waves recorded on a fiber-optic cable can localize groundwater-driven S-wave velocity changes at meter scale and daily cadence.

desk verdict A reproducible two-year vehicle-DAS FWI groundwater monitoring study with a genuine seasonal and lateral signal, but the 'daily' and 'single-day' claims are not supported by the stated 14-day moving average. read the letter →

arxiv 2501.10618 v1 pith:3REGVCV5 submitted 2025-01-18 physics.geo-ph

Haipeng Li , Jingxiao Liu , Shujuan Mao , Siyuan Yuan , Robert G. Clapp , Biondo L. Biondi This is my paper · ORCID
classification physics.geo-ph
keywords groundwatermonitoringdistributedacousticsensingfull-waveforminversionvehicle-inducedseismicsourcestime-lapseS-wavevelocityporoelasticityaquiferrecharge
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

Using two years of continuous recordings on a 1 km fiber-optic cable along Sandhill Road, California, this paper tries to show that elastic full-waveform inversion of virtual source gathers generated by ordinary vehicular traffic can monitor groundwater dynamics on a daily basis. The key result is a set of daily S-wave velocity models in which a 2.9% velocity reduction tracks a 9.0-meter rise in the water table during Water Year 2023, with the largest changes concentrated at 15–30 m depth beneath grassy areas. If correct, the method would give aquifer managers meter-scale spatial localization and daily temporal sampling without active seismic sources, using only a pre-existing telecommunications cable and the traffic already on the road. The paper also argues that the observed velocity drops are poroelastic: rising pore pressure lowers effective stress and hence S-wave velocity, matching a coupled flow-mechanics simulation.

What carries the argument

The load-bearing machinery is the combination of targeted interferometry and time-lapse elastic full-waveform inversion. Targeted interferometry tracks individual vehicles from the quasi-static deformation signals in the DAS record, selects coherent surface-wave windows, cross-correlates them, and stacks the results into daily virtual source gathers with 39 virtual sources and receivers spaced every 8.16 m. The inversion builds a baseline model with a multi-scale full-waveform inversion using a global-correlation objective function, then monitors changes with a modified double-difference strategy that inverts differences between baseline and monitor waveforms while normalizing amplitudes. DAS acquisition is modeled as averaged tangential strain over the 16 m gauge length, and S-wave velocity is the only parameter inverted, with P-wave velocity and density coupled through an empirical velocity ratio and density–velocity relation. The physical link to groundwater is the effective-stress relation $P_{\text{eff}} = P_{\text{conf}} - P_{\text{pore}} + S_w\psi$, by which rising pore pressure reduces S-wave velocity; pore-pressure time series are generated with a diffusive poroelastic solution for surface loading.

What would settle it

Rerun the same daily time-lapse full-waveform inversion without the 14-day moving average and inspect the velocity-change time series around December 31, 2022: if no sharp single-day step remains, the claimed daily-resolution detection is an artifact of smoothing. A complementary test would compare daily FWI velocity changes against a continuous pressure-transducer water-level record during a later atmospheric-river event to see whether the seismic response really is same-day.

Watch

Extended reading notes

Core claim

The central discovery claimed is that time-lapse elastic full-waveform inversion applied to daily stacked virtual source gathers from vehicle-induced surface waves resolves absolute, localized S-wave velocity changes tied to aquifer storage. Over the monitoring period the inversion recovers a 2.9% reduction in S-wave velocity during the wet Water Year 2023, coincident with a cumulative 9.0-meter groundwater-table rise, and attributes a rapid velocity change on the day of the December 31, 2022 extreme precipitation event to a fast water-table rise that monthly well measurements missed. Velocity changes appear mainly below the local water table, at 15–30 m depth, are stronger under grassy areas than under a paved parking lot, and correlate with pore-pressure increases simulated from precipitation records. The paper presents this as an advance beyond coda-wave monitoring because full-waveform inversion localizes changes in absolute velocity rather than giving spatially averaged relative values.

Load-bearing premise

The claim of daily resolution rests on the assumption that the 14-day moving window average applied to the daily virtual-source gathers to boost signal-to-noise does not blur away the single-day velocity change attributed to the December 31, 2022 storm; if the smoothing is what makes the gathers usable, the one-day attribution cannot be recovered from those data.

Editorial extensions

If this is right

  • Daily FWI of vehicle-DAS data localizes S-wave velocity changes in depth and along the cable, so groundwater response can be mapped where monitoring wells are too sparse and coda-wave methods give only spatially averaged relative changes.
  • The same data stream detects a single-day aquifer response to the December 31, 2022 extreme precipitation event, something the monthly hydraulic-head measurements from the nearby well missed.
  • Because the sources are ordinary vehicles and the sensors are pre-existing telecommunications fibers, the approach can replace costly daily active-source seismic surveys for near-surface aquifer monitoring.
  • The observed contrast between grassy and paved areas implies that impervious surfaces suppress local recharge, providing a direct geophysical measure of urbanization effects on aquifer recharge.
  • Extending the approach to fiber-optic networks with 2-D geometries would support 3-D aquifer imaging at daily cadence.

Reading between the lines

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

  • Inference: if the 14-day moving average is required for stable virtual source gathers, the practically supported temporal resolution may be closer to two weeks than one day; a no-smoothing test would settle which claim survives.
  • Inference: pairing daily FWI with continuous pressure-transducer water levels at one or more wells would calibrate the velocity-change-to-storage relation, turning the method from a qualitative proxy into a quantitative head or storage estimator.
  • Inference: transfer to other urban fiber segments will depend on traffic density and low-frequency source content; heavy-truck traffic should reach deeper and stabilize daily gathers, while quiet residential segments may need longer stacking windows.
  • Inference: because the paper notes that saturation changes above the water table are invisible to S-wave surface-wave inversion, adding P-wave sensitivity to the same DAS data could extend monitoring into the vadose zone.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper presents a 27-month (December 2021 to March 2024) case study along Sandhill Road, California, in which virtual source gathers (VSGs) retrieved from vehicle-induced DAS signals by targeted interferometry are inverted with time-lapse elastic full-waveform inversion (FWI) to image S-wave velocity changes. The authors report a 2.9% Vs reduction during Water Year 2023, corresponding to a roughly 9 m water-table rise, lateral differences between grassy and paved areas, and a claimed one-day response to a December 31, 2022 extreme precipitation event. They validate the velocity changes against well measurements and precipitation records, and support the interpretation with a 1D poroelastic pore-pressure model. The paper also shares the targeted-interferometry code, daily FWI models, and selected VSGs via GitHub and Zenodo.

Significance. If the claims hold, this work is significant: it demonstrates that non-uniform vehicle traffic recorded on dark fiber can be converted into repeatable virtual-source surveys and inverted with FWI at a cadence and cost far below active-source monitoring, while providing lateral and depth localization that coda-wave monitoring cannot. The study is strengthened by the use of independent well and precipitation data, checkerboard resolution tests, and a poroelastic interpretation that is not itself used to construct the velocity-change field. The open release of code, daily FWI models, and selected VSGs is a further strength. However, the central novelty is the claimed daily temporal resolution, and that claim is not supported by the processing description as written.

major comments (3)
  1. [Materials and Methods, Time-lapse targeted interferometry; Results, Impact of Extreme Precipitation Events] The Methods state that 'a 14-day moving window average was applied to the daily data to enhance SNR further.' The Results (Fig. 4d) and the accompanying text attribute a rapid groundwater-table rise to the December 31, 2022 EPE 'within one single day.' A 14-day moving average has an effective temporal resolution of roughly 14 days (half-width 7 days, or 13 days if trailing), so a single-day attribution is not recoverable from the smoothed input unless the smoothing is applied only to display curves, which the text does not say. Please clarify the role of the smoothing in the FWI workflow and, if it is part of the data preprocessing, either re-invert unsmoothed daily VSGs or revise the abstract, title, and temporal-resolution claims (including the 'daily analysis' and the >6-month recession-time estimate) to a ~14-day effective resolution.
  2. [Materials and Methods, Time-lapse targeted interferometry (trace rejection); Supplementary Fig. S2] The quality-control criteria remove monitor traces with time delays exceeding 25 ms and mute traces with NRMS greater than 0.6. Supplementary Fig. S2 shows that NRMS systematically increases during wet periods, so these criteria preferentially reject exactly the traces that carry the largest hydrological signal. This selection can bias the inverted velocity-reduction amplitudes (most plausibly downward) and can distort the grassy-versus-paved comparison if the rejection rate differs by surface type. Please report the fraction of rejected or muted traces as a function of date and location, and provide a sensitivity test of the 2.9% reduction and the spatial patterns in Fig. 3a and Fig. 4a to the 25 ms and 0.6 thresholds.
  3. [Discussion, final paragraph; Supplementary Fig. S5] The Discussion claims 'meter-scale spatial resolution,' but the checkerboard resolution tests in Supplementary Fig. S5 recover anomalies of 25 m depth by 50 m lateral and 20 m by 40 m, and the Introduction states 'tens-of-meter resolution in space.' The 'meter-scale' phrasing is inconsistent with the paper's own resolution analysis and should be corrected to the demonstrated resolution.
minor comments (5)
  1. [Materials and Methods, Eq. (8)] Equation (8) contains typesetting errors in the erf/erfc arguments (the square-root symbols and fractions are mis-rendered); please ensure the formula is typeset correctly.
  2. [Supplementary Fig. S2 caption] The caption refers to a 'virtual source at 520 m' while the main text describes the same location as a 'virtual receiver at 520 m'; please clarify which is intended.
  3. [Data and materials availability] The released continuous DAS recordings are described as 'from December 2022'; please clarify the exact date range of the released data, given the full two-year monitoring period.
  4. [Supplementary Fig. S5 caption] The word 'checkboard' in the caption should be 'checkerboard.'
  5. [Materials and Methods, Time-lapse targeted interferometry] Please specify whether the 14-day moving average is centered or trailing, and state whether the baseline period of July 7 to August 25, 2022 is also smoothed before being used as the reference.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the FWI velocity-change field is data-derived, and the independent well/precipitation comparisons are validation, not fitted inputs.

full rationale

The derivation chain is self-contained. The baseline model is produced by elastic FWI of the retrieved VSGs using a global-correlation objective and multi-scale band inversion, and the time-lapse velocity changes are produced by a double-difference global-correlation FWI of monitor VSGs relative to the baseline (Eqs. 4-7). The velocity-change field is therefore an inversion output, not a quantity read off from any fitted parameter; no model parameter is calibrated to the well or precipitation data and then renamed a prediction. The independent constraints (well 06S03W02D032 water levels, KCAMENLO88 precipitation, prior SASW dispersion results) are used only for comparison and validation, not as inputs to the inversion. The poroelastic pore-pressure calculation (Eqs. 8-9) uses a separate precipitation record with stated constants (hydraulic diffusivity 0.1, Skempton coefficient 0.8, undrained Poisson ratio 0.4); those constants are not fit to the observed velocity changes, and the modeled pressure is presented after the inversion as an interpretive comparison (Fig. 4b), so it cannot be a self-definitional input. The main self-citation is the targeted interferometry workflow (ref. 30, Yuan et al. 2024), a published method with public code (github.com/syyuan93/das_veh) and is demonstrated here on the Stanford DAS-2 data (e.g., VSG stacked from 125 vehicles on Dec 3, 2022); it is not invoked as a uniqueness theorem or as justification of the FWI result. One non-circular caveat: the Methods state 'a 14-day moving window average was applied to the daily data to enhance SNR further'; this temporal filtering potentially undercuts the literal 'within one single day' attribution for the Dec 31, 2022 EPE, but it is a temporal-resolution concern, not an equivalence-by-construction of the measured velocity change to its input. No equation in the paper reduces by construction to a fitted value or to a self-citation, so no circular step is identified.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central FWI result stands on several domain assumptions: a 2D elastic model along the cable, S-wave-only inversion with fixed Vp/density relations, and repeatable traffic-DAS sources. The causal interpretation additionally assumes pore-pressure dominance over thermal and moisture effects, and the poroelastic comparison rests on three hand-set parameters. No new physical entities are introduced.

free parameters (5)
  • hydraulic diffusivity c = 0.1
    Used in the Roeloffs pore-pressure solution (Eq. 9); no uncertainty or sensitivity analysis is provided. The simulated pore pressure time series is compared directly to observed velocity changes.
  • Skempton's coefficient B = 0.8
    Chosen in the pore-pressure model; B scales the undrained response in Eqs. 8-9 and therefore the amplitude and timing of predicted pore pressure.
  • undrained Poisson's ratio nu_u = 0.4
    Chosen in the pore-pressure model; it enters the elastic constant alpha in Eq. 8.
  • Vp/Vs ratio = 1.732
    P-wave velocity is set as 1.732 times S-wave velocity rather than inverted; this fixed empirical scaling affects synthetic waveforms and inversion coupling.
  • Gardner's density relation = Gardner empirical relation (coefficients not stated)
    Density is not inverted; it is derived from Vp via Gardner's relationship, adding another fixed empirical assumption to the inversion.
assumptions (5)
  • domain assumption The observed velocity changes are caused primarily by pore-pressure changes in the saturated zone; thermal, barometric, and soil-moisture effects are secondary.
    The paper states the changes are 'likely driven primarily by the shallow sediment's poroelastic response' and acknowledges thermal and moisture complications in the Discussion; the groundwater interpretation depends on this dominance.
  • domain assumption A 2D elastic model along the fiber is sufficient to represent the recorded wavefield; out-of-plane heterogeneity and traffic source location errors are negligible.
    The FWI solves the 2D stress-velocity equations for a line source with a free surface; the urban environment contains 3D structures and moving sources.
  • domain assumption The empirical Vp/Vs ratio and Gardner density scaling are adequate to model Rayleigh-wave propagation for this site.
    Only S-wave velocity is inverted; P-wave velocity and density are fixed. If actual Vp or density deviate, the S-wave update will absorb those errors.
  • standard math Finite-difference wave propagation, DAS gauge-length averaging, and the global-correlation objective function are implemented correctly.
    The inversion relies on standard numerical methods and published adjoint-state theory; no formal verification is provided.
  • domain assumption Hydraulic well 06S03W02D032 and weather station KCAMENLO88 represent conditions along the 1 km fiber.
    These point measurements are used to validate the spatially distributed FWI results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Daily Groundwater Monitoring Using Vehicle-DAS Elastic Full-waveform Inversion." pith.science (2026). https://pith.science/paper/3REGVCV5

@misc{pith2026250110618,
  author       = {Pith},
  title        = {Pith review of: Daily Groundwater Monitoring Using Vehicle-DAS Elastic Full-waveform Inversion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3REGVCV5}},
  note         = {Machine review of arXiv:2501.10618}
}
read the original abstract

Understanding groundwater dynamics is critical for sustainable water management, particularly as climate extremes intensify. However, the resolutions of existing subsurface observational tools are still inadequate for detailed aquifer monitoring and imaging. We introduce an innovative technique for groundwater monitoring using time-lapse full-waveform inversion, leveraging fiber-optic cables as seismic sensors and vehicular traffic as repetitive seismic sources. Over a two-year period along Sandhill Road, California, this approach captures detailed spatiotemporal S-wave velocity variations, revealing a 2.9% reduction corresponding to a 9.0-meter groundwater table rise after atmospheric-river storms in Water Year 2023. Notably, this approach enables the high-resolution daily analysis of rapid aquifer responses. We observe spatially inhomogeneous velocity changes, with less reduction beneath impervious paved zones than under grassy areas, underscoring the impact of urbanization on the natural recharge of aquifers. Our findings highlight the potential of Vehicle-DAS FWI for high-resolution daily monitoring and quantitative spatiotemporal characterizations of groundwater systems.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Real-time processing of distributed acoustic sensing data for earthquake monitoring operations

    physics.geo-ph 2025-05 conditional novelty 6.0 of 10

    A modular Python framework integrates real-time distributed acoustic sensing (DAS) data and machine-learning phase picks into the operational AQMS earthquake monitoring system, demonstrated on a 100-km Ridgecrest array.

Reference graph

Works this paper leans on

4 extracted references · 3 canonical work pages · cited by 1 Pith paper

  1. [11]

    Characterizing Vehicle-Induced Distributed Acoustic Sensing Signals for Accurate Urban Near-Surface Imaging

    N. Houlie´, G. J. Funning, R. Bu¨rgmann, Use of a GPS-derived troposphere model to improve InSAR deformation estimates in the San Gabriel Valley, California. IEEE Transactions on Geoscience and Remote Sensing 54 (9), 5365–5374 (2016). 12. A. Rezaei, Z. Mousavi, Characterization of land deformation, hydraulic head, and aquifer properties of the Gorgan conf...

    work page Pith review arXiv 2016

  2. [40]

    Y. Pan, L. Gao, T. Bohlen, High-resolution characterization of near-surface structures by surface-wave inversions: from dispersion curve to full waveform. Surveys in Geophysics 40, 167–195 (2019). 41. I. Wong, et al., Seismic microzonation and development of seismic design ground motions for the Stanford University campus, California, in Proceedings of th...

    work page 2019

  3. [56]

    J. Liu, S. Yuan, Y. Dong, B. Biondi, H. Y. Noh, TelecomTM: A Fine-Grained and Ubiquitous Traffic Monitoring System Using Pre-Existing Telecommunication Fiber-Optic Cables as Sensors. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 7 (2), 1–24 (2023). 57. E. Kragh, P. Christie, Seismic repeatability, normalized rms, and ...

    2023

  4. [71]

    E. A. Roeloffs, Fault stability changes induced beneath a reservoir with cyclic variations in water level. Journal of Geophysical Research: Solid Earth 93 (B3), 2107–2124 (1988). 72. P. Segall, Earthquake and volcano deformation (Princeton University Press) (2010). 73. D. Rivet, F. Brenguier, F. Cappa, Improved detection of preeruptive seismic velocity dr...

    work page 1988

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.