Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Video-based Direct Time Series Measurement of Along-Strike Slip on the Coseismic Surface Rupture During the 2025 Mw7.7 Myanmar Earthquake

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read This paper presents the first in-situ, high-rate direct measurement of coseismic slip evolution, recovered from CCTV footage of the 2025 Mw 7.7 Myanmar earthquake, and derives a critical slip-weakening distance from it.

desk verdict A genuinely new dataset—a 30 Hz video record of coseismic surface slip—with a solid pixel-tracking pipeline, but the physical scale is uncertain by a factor of ~1.6–1.8 and the Dc value is only as good as that scale. read the letter →

arxiv 2505.20494 v1 pith:6ZTBQ3GZ submitted 2025-05-26 physics.geo-ph

classification physics.geo-ph
keywords coseismicslipsurfaceruptureSagaingFaultCCTVvideocomputervisiontrackingslip-weakeningdistancedisplacementtimeseries2025Myanmarearthquake
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

This paper argues that a publicly available CCTV video, recorded about 15 meters from the surface trace of the Sagaing Fault during the 2025 Mw 7.7 Myanmar earthquake, preserves the full sub-second evolution of along-strike surface slip. By stabilizing and undistorting the frames, tracking reference objects with two computer-vision methods and manual checking, then calibrating pixel motion with on-site measurements of plant pots, a fence grid, and a curb, the authors recover a 30-Hz physical displacement time series. The final offset at the reference object is about 280 cm, while a concrete curb gives a lower bound of 193 cm. From the smoothed displacement, they estimate a critical slip-weakening distance $D_c \approx 1.67$–$1.68$ m over a duration of $0.77$ s. If correct, this is the first direct, high-rate record of fault slip during an earthquake and a benchmark for dynamic rupture and friction models.

What carries the argument

The load-bearing machinery is a calibration chain from 30 fps CCTV pixels to physical ground displacement. Four stable regions (two ground lights and two walls) are used to estimate affine transforms that stabilize the image against camera shake; radial distortion is corrected with a barrel model using $k_1=-0.06$, $k_2=0.002$, $k_3=0$. Pixel trajectories are then obtained with grayscale normalized cross-correlation template matching and with CoTracker, a transformer-based multi-point tracker, and are verified against manual annotations. Physical scale comes from field measurements: the distance between two plant pots (440 cm) maps the final pixel offset to 280 cm, a similar-triangles projection off the fence bar grid gives 287 cm, and the concrete curb gives a lower-bound 193 cm. The displacement time series is smoothed with a Savitzky-Golay filter, differentiated to velocity, and integrated via Simpson's rule between the rupture onset (minimum displacement) and peak velocity to obtain $D_c$.

What would settle it

Measure the actual present-day distance between the two plant pots with a tape and compare it with the value used here (440 cm), and independently survey the permanent offset of the concrete curb or another rigid marker at the same location. If the plant-pot distance has changed, or if a GPS or total-station survey gives a final offset near 4–5 m instead of 2.8 m, the time series and the derived $D_c \approx 1.67$ m would need to be rescaled and would not be a direct measure of fault slip as stated.

Watch

Extended reading notes

Core claim

The central discovery is a time-resolved record of coseismic surface slip measured directly at the fault, not inverted from seismic or geodetic data. Using grayscale template matching and a transformer-based tracker (CoTracker), the authors obtain consistent pixel trajectories that are also reproduced by manual frame-by-frame tracking. After conversion to physical units via the known separation of two plant pots (measured today as 440 cm), the x-displacement series reaches about 280 cm; the same conversion applied to a solar-panel corner viewed through a fence yields about 287 cm, and a field-measured curb offset gives a smaller 193 cm lower bound. The final displacement time series shows a brief anti-slip phase followed by rapid acceleration, and differentiating it yields a velocity peak and an estimated critical slip-weakening distance $D_c \approx 1.68$ m over $0.77$ s. The authors provide the CoTracker pixel time series in an appendix so others can rescale it to physical units.

Load-bearing premise

The entire scale of the displacement time series rests on the assumption that the 440 cm distance between the two plant pots measured after the earthquake is close to what it was at the moment of rupture; if the true scale were closer to the 4–5 m suggested by satellite offsets, the displacement values and $D_c$ would roughly double.

Editorial extensions

If this is right

  • The published 30 Hz pixel time series (Appendix II) gives dynamic rupture modelers a direct, sub-second observation of on-fault slip to test against simulated slip histories.
  • The estimated $D_c \approx 1.7$ m and slip duration of $0.77$ s become concrete targets for rate-and-state and slip-weakening friction laws in near-field ground-motion simulations.
  • The observed anti-slip phase before the main rupture provides a check on models of pre-rupture seismic-wave loading and fault reversal.
  • Cross-validation of template matching, CoTracker, and manual tracking suggests the pixel kinematics are robust, which supports using CCTV archives to recover slip time series in future earthquakes.

Reading between the lines

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

  • If this approach generalizes, dense CCTV networks near active faults could turn ordinary surveillance footage into a new class of near-field seismic data, complementing strong-motion and geodetic observations.
  • The local 2.8 m surface slip being smaller than the 4–5 m satellite offset suggests surface rupture may underestimate fault-zone slip; the time series could be combined with inversion results to quantify shallow slip deficit.
  • A testable extension would be to apply the same stabilization-and-tracking pipeline to other accidentally captured earthquake videos to see whether the short anti-slip phase is a common pre-rupture feature.
  • The paper's uncertainty is dominated by the single reference scale; using multiple independent field references or an onboard scale bar could reduce that uncertainty in future applications.
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.

Referee Report

3 major / 5 minor

Summary. This manuscript analyzes a publicly available CCTV recording that captured the surface rupture of the 2025 Mw 7.7 Sagaing Fault earthquake near Mandalay, Myanmar. The authors stabilize and undistort the video, track pixel displacements of objects near the fault using template matching and CoTracker, validate both against manual tracking, and convert pixel displacements to physical displacements using field-measured reference objects. From the resulting time series they estimate a critical slip-weakening distance of Dc ≈ 1.67–1.68 m. The appendix provides the full CoTracker pixel time series.

Significance. If the scale calibration is accepted, this is a potentially first-of-its-kind direct, high-sampling-rate (30 Hz) in-situ time series of coseismic surface slip, with genuine value for validating dynamic rupture simulations and friction models. The paper's strengths include the multi-method cross-validation of pixel tracking (template matching, CoTracker, and manual), the openly available video source, and the complete pixel time series in Appendix II. The authors are also transparent about several limitations. The main weakness is that the physical scale rests on an unverified assumption about a reference-object distance, and the paper's own Discussion acknowledges a factor-of-1.6–1.8 discrepancy with satellite-derived offsets, so the headline numerical results are contingent on resolving that calibration issue.

major comments (3)
  1. [Section 3.2.1, Appendix I] The 280 cm normalization factor rests on the assumption that the current pot-to-pot distance D1 = 440 cm equals the distance at the time of the earthquake; Section 3.2.1 concedes this is not guaranteed. Because the physical displacement time series and the Dc estimate scale linearly with this factor, an alternative scale of 4–5 m suggested by satellite feature tracking in the Discussion would change Dc from ≈1.67 m to roughly 2.7–3.0 m. Please provide a quantitative bound or independent verification of D1 at the time of the earthquake, present an uncertainty range for the normalization factor, and show how the displacement time series and Dc would change if the larger scale were used.
  2. [Section 2.3, Section 3.2.2] The camera projection and scale model is underconstrained: the FOV is fixed to 90° and k1 = −0.06 by visual alignment of vanishing points, and the fence-based estimate of 287 cm invokes 'known distances between the camera and the fence, and between the fence and the target objects' that are not reported in the manuscript. Please report those distances, validate the projection against measured reference dimensions (e.g., the gate width W3-1 = 740 cm), and include a sensitivity analysis of the inferred physical slip to the FOV, distortion coefficients, and the assumed camera-to-object geometry.
  3. [Section 3.3, Figure 15] The reported Dc values depend on an ad hoc definition of the dynamic-rupture window (minimum displacement to velocity peak) and on Savitzky-Golay smoothing parameters, but no uncertainty quantification is provided. Because Dc is a headline quantitative result, the manuscript should test how the estimate varies with the smoothing window and with the chosen rupture-onset and end-of-weakening criteria, and should report the resulting spread alongside the Dc values.
minor comments (5)
  1. [Section 2.2] The phrase 'we took below pixel regions as reference' should read 'the following pixel regions were used as reference.'
  2. [Figure 4] The chosen distortion parameter set (FOV = 90°, k1 = −0.06) should be explicitly marked in Figure 4, since the reader cannot otherwise identify which panel corresponds to the selected parameters.
  3. [Section 3.1, Section 4] The relationship between CCTV watermark time and Myanmar Standard Time should be stated once and explicitly; the text notes that the two differ and that the video is about 4 minutes behind MST, but does not clearly state the offset between the watermark times in the figures and the origin time of the earthquake.
  4. [Figure 13] The y-axis label 'Real-work displacement' should be 'Real-world displacement.'
  5. [Appendix II] The appendix table should include a clear statement of the time origin (12:46:31.666 watermark time) and note that x and y are sub-pixel CoTracker values, so that users do not mistake them for integer pixel coordinates.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the displacement measurement and Dc estimate are derived from independent pixel tracking and field calibration, with acknowledged scale uncertainty.

full rationale

The central claim is a direct observational measurement, not a prediction from a model. The pixel displacement time series is produced by three independent tracking methods (template matching, CoTracker, manual) whose agreement is shown in Figures 6-8. Conversion to physical units uses field-measured reference objects: the plant-pot spacing (D1=440 cm) and a fence-grid projection (approx. 287 cm), giving roughly 280 cm of total slip. The paper explicitly states the pot-distance assumption ('the current distance between the two plant pots is not guaranteed to be identical to that pair at the time of the earthquake, but is assumed to be close'), which is an acknowledged uncertainty rather than a circular input. The Dc values (1.67 m and 1.68 m) are computed by integrating the measured velocity between a defined onset (minimum displacement) and end (peak velocity); this is an estimation from the recorded time series, not a prediction that feeds back into the measurement. No load-bearing self-citations, uniqueness theorems, or ansatz-smuggling citations appear; the cited references are external tools (OpenCV, CoTracker) or independent observations (Turkey 2023, Latour et al. 2025). The discrepancy with satellite-derived 4-5 m slip is discussed as a possible scale or local-effect issue, but this concerns accuracy, not circularity. The paper would be strengthened by additional validation of the scale, but no step in the derivation reduces to its own output.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities or forces. Its free parameters are mainly processing choices (distortion coefficients, search window, stabilization regions) and the key normalization factor of 280 cm that converts pixels to physical slip. The assumptions concern camera geometry, stabilization reference stability, and rigid-block ground motion. No objects beyond standard video-processing concepts are postulated.

free parameters (5)
  • Distortion coefficients k1, k2, k3 and FOV = k1=-0.06, k2=0.002, k3=0, FOV=90 degrees
    Chosen by visually aligning vanishing points between the first and last frames (Section 2.3). The paper states the impact on displacement is limited, but it is an unconstrained choice.
  • Template matching search window = 4 pixels
    A hyperparameter restricting the search region to improve robustness (Section 2.4).
  • Stabilization reference regions = Four rectangles at (344,398,29,99), (1152,335,19,68), (333,16,121,146), (790,145,175,182)
    Ad hoc selections of assumed stable objects (two ground lights and two walls) used to estimate the affine transform. If any of these moved with the fault, the stabilization would be biased.
  • Dc window = Onset at minimum displacement, end at peak velocity
    The integration interval for Dc is defined by features of the same signal being analyzed, making the result dependent on this choice (Section 3.3).
  • Normalization factor = 280 cm
    Field-derived offset that scales pixel displacements to physical meters (Section 3.2.4). Its uncertainty (a factor of about 2 relative to satellite estimates) directly scales all physical results including Dc.
assumptions (4)
  • domain assumption The image x-axis is aligned with the fault-parallel (north-south) direction of slip.
    The paper uses x-displacement as the along-strike slip without a quantitative camera calibration to the fault trace orientation. This enters in Section 3.2.4 when converting pixel x motion to physical fault-parallel displacement.
  • domain assumption The four stabilization reference regions are stationary relative to the camera during the earthquake.
    The affine stabilization assumes these objects do not move with the fault. If any reference region lies on the moving block, true slip would be removed. This is used in Section 2.2.
  • domain assumption The OpenCV radial distortion model adequately describes the CCTV lens, with principal point at the image center (640, 360).
    The distortion correction uses this model and a 90-degree FOV. The paper acknowledges other parameter sets could give similar results. Appears in Section 2.3.
  • domain assumption The ground deformation is rigid-body motion of discrete blocks; objects in the scene move without local deformation or rotation beyond what is tracked as pixel translation.
    The affine stabilization and point tracking assume planar rigid motion of the visible objects. The paper notes some regions experienced uplift and anti-slip deformation, indicating this is only an approximation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Video-based Direct Time Series Measurement of Along-Strike Slip on the Coseismic Surface Rupture During the 2025 Mw7.7 Myanmar Earthquake." pith.science (2026). https://pith.science/paper/6ZTBQ3GZ

@misc{pith2026250520494,
  author       = {Pith},
  title        = {Pith review of: Video-based Direct Time Series Measurement of Along-Strike Slip on the Coseismic Surface Rupture During the 2025 Mw7.7 Myanmar Earthquake},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6ZTBQ3GZ}},
  note         = {Machine review of arXiv:2505.20494}
}
read the original abstract

This study presents a time-resolved analysis of coseismic lateral surface rupture along the Sagaing Fault during the Mw 7.7 Mandalay, Myanmar earthquake on March 28, 2025. Leveraging a publicly available Closed-Circuit Television (CCTV) footage alongside on-site measurements, we show the first in-situ high sampling rate direct measurement of a coseismic slip evolution of a fault during an earthquake. Our work comprises four primary stages: data acquisition, video pre-processing, object tracking, and physical displacement estimation. Video pre-processing includes camera stabilization and distortion correction. We then track pixel-level movements of selected reference points using two complementary computer-vision approaches -- a traditional grayscale template matching algorithm and a state-of-the-art vision transformer multi-object tracking algorithm, and verify both profiles against meticulous manual frame-by-frame measurements, with results that closely match one another. Finally, we translated those pixel displacements into real-world ground movements by calibrating against reference objects whose dimensions were measured on site. Based on the resulting displacement time series, we estimated the critical slip-weakening distance. The resulting high-resolution time series of the along-strike slip, provided in the appendix, offers a critical benchmark for validating dynamic rupture simulations, refining frictional models, and enhancing seismic hazard assessment.

Figures

Figures reproduced from arXiv: 2505.20494 by the authors.

Figure 1
Figure 1. Location of the surveillance camera (yellow circle) and the fault trace (red line) ruptured [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Four selected regions were used for video stabilization to correct camera shake. Best [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Frames after applying stabilization and distortion correction. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (16 more)
Figure 4
Figure 4. Figure 4: Some other results with different sets of parameters. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Selected Targets. The results presented in [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Pixel displacement comparison: Template matching vs. Manual tracking. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Pixel displacement comparison: CoTracker vs. Manual tracking. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Pixel displacement of object 7. 9 [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Displacement and physical dimensions of the target object. [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Tracking solar panel behind fence [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: Estimation from fence bar gap. This diagram was created using Tinkercad, a free online [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 12
Figure 12. Figure 12: Direct Measurement of Fault Relative Displacement. [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 13
Figure 13. Figure 13: Real-work displacement estimation of the plant pot. [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]
Figure 14
Figure 14. Figure 14: Smoothed displacement time series derived from template matching and CoTracker [PITH_FULL_IMAGE:figures/full_fig_p014_14.png]
Figure 15
Figure 15. Figure 15: Velocity profiles and estimated critical slip-weakening distances ( [PITH_FULL_IMAGE:figures/full_fig_p014_15.png]
Figure 16
Figure 16. Figure 16: Comparison of tracking results between the original video and the calibrated video. [PITH_FULL_IMAGE:figures/full_fig_p015_16.png]
Figure 17
Figure 17. Figure 17: Deformation occurs in the anti-slip direction prior to the slip, and some regions suffer [PITH_FULL_IMAGE:figures/full_fig_p016_17.png]
Figure 18
Figure 18. Figure 18: Reference Object Size for Video Analysis. Measured in cm. [PITH_FULL_IMAGE:figures/full_fig_p018_18.png]
Figure 19
Figure 19. Figure 19: Direct observation of fault displacement after earthquake. [PITH_FULL_IMAGE:figures/full_fig_p019_19.png]

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. Supershear-subshear-supershear rupture sequence during the 2025 Mandalay Earthquake in Myanmar

    physics.geo-ph 2025-06 conditional novelty 6.0 of 10

    The 2025 Mandalay earthquake rupture went supershear (~6 km/s), slowed to subshear (~3 km/s) before the video site, then re-accelerated to supershear.

Reference graph

Works this paper leans on

14 extracted references · 10 canonical work pages · cited by 1 Pith paper

  1. [1]

    The OpenCV Library

    G. Bradski. “The OpenCV Library”. In:Dr. Dobb’s Journal of Software Tools(2000)

  2. [2]

    CoTracker: It is Better to Track Together

    Nikita Karaev et al. “CoTracker: It is Better to Track Together”. In:Computer Vision – ECCV 2024. Springer, 2024, pp. 18–35.doi: 10.1007/978-3-031-73033-7_2 . url: https: //doi.org/10.1007/978-3-031-73033-7_2

  3. [3]

    Nikita Karaev et al.CoTracker3: Simpler and Better Point Tracking by Pseudo-Labelling Real Videos. 2024. url: https://arxiv.org/abs/2410.11831

  4. [4]

    Slip history of the 2016 Mw 7.0 Kumamoto earth- quake: Intraplate rupture in complex tectonic environment

    Jinlai Hao, Chen Ji, and Zhenxin Yao. “Slip history of the 2016 Mw 7.0 Kumamoto earth- quake: Intraplate rupture in complex tectonic environment”. In:Geophysical Research Letters 44.2 (2017), pp. 743–750

  5. [5]

    Rupture characteristics of major and great (Mw≥ 7.0) megathrust earth- quakes from 1990 to 2015: 1. Source parameter scaling relationships

    Lingling Ye et al. “Rupture characteristics of major and great (Mw≥ 7.0) megathrust earth- quakes from 1990 to 2015: 1. Source parameter scaling relationships”. In:Journal of Geo- physical Research: Solid Earth121.2 (2016), pp. 826–844

  6. [6]

    Rupture directivity of the 2021 MW 6.0 Yangbi, Yunnan earthquake

    Wenzheng Gong et al. “Rupture directivity of the 2021 MW 6.0 Yangbi, Yunnan earthquake”. In: Journal of Geophysical Research: Solid Earth127.9 (2022), e2022JB024321

  7. [7]

    Coseismic slip model of the 2008 Wenchuan earthquake derived from joint inversion of interferometric synthetic aperture radar, GPS, and field data

    Xiaopeng Tong, David T Sandwell, and Yuri Fialko. “Coseismic slip model of the 2008 Wenchuan earthquake derived from joint inversion of interferometric synthetic aperture radar, GPS, and field data”. In:Journal of Geophysical Research: Solid Earth115.B4 (2010)

  8. [8]

    Rupture process of the 2021 M7. 4 Maduo earthquake and implication for deformation mode of the Songpan-Ganzi terrane in Tibetan Plateau

    Han Yue et al. “Rupture process of the 2021 M7. 4 Maduo earthquake and implication for deformation mode of the Songpan-Ganzi terrane in Tibetan Plateau”. In:Proceedings of the National Academy of Sciences119.23 (2022), e2116445119

Show all 14 references
  1. [9]

    Zenodo, Apr

    Maximillian van Wyk de Vries.2025 Myanmar earthquake displacement maps. Zenodo, Apr

  2. [10]

    Ground surface displacement maps of the 2025 Myanmar earthquake

    Simon Gascoin. Ground surface displacement maps of the 2025 Myanmar earthquake. Zenodo, Apr. 2025. doi: 10.5281/zenodo.15127866 . url: https://doi.org/10.5281/zenodo. 15127866

  3. [11]

    A simple method for reconstructing a high-quality NDVI time-series data set based on the Savitzky–Golay filter

    Jin Chen et al. “A simple method for reconstructing a high-quality NDVI time-series data set based on the Savitzky–Golay filter”. In:Remote sensing of Environment91.3-4 (2004), pp. 332–344

  4. [12]

    The sharp turn: Backward rupture branching during the 2023 Mw 7.8 Kahramanmaraş (Türkiye) earthquake

    Xiaotian Ding et al. “The sharp turn: Backward rupture branching during the 2023 Mw 7.8 Kahramanmaraş (Türkiye) earthquake”. In:Seismica 2.3 (2023)

  5. [13]

    Direct Estimation of Earthquake Source Properties from a Single CCTV Camera

    SoumayaLatouretal. Direct Estimation of Earthquake Source Properties from a Single CCTV Camera. 2025. arXiv:2505.15461 [physics.geo-ph]. url: https://arxiv.org/abs/2505. 15461. 22

  6. [2025]

    url: https://doi.org/10.5281/zenodo.15123647

    doi: 10.5281/zenodo.15123647. url: https://doi.org/10.5281/zenodo.15123647

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.