Pith. sign in

REVIEW 5 major objections 6 minor 9 references

Wavefront Profiling via correlation of GLAO open loop telemetry

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Spatio-temporal cross-correlations of open-loop wavefront sensor telemetry from the 'imaka ground-layer adaptive optics system recover turbulent layer velocities that broadly agree with independent summit weather measurements.

desk verdict Useful application of an established wind-profiling technique to 'imaka telemetry, but the claimed agreement with external winds is undermined by a systematic 45-degree direction offset the paper neither explains nor quantifies. read the letter →

arxiv 2608.10786 v1 pith:2NCKIJGZ submitted 2026-08-11 astro-ph.IM

Eden A. McEwen , Ryan Dungee , Mark Chun , Jessica R. Lu , Olivier Lai , Christoph Baranec This is my paper · ORCID
classification astro-ph.IM
keywords AdaptiveOpticsOpticalturbulenceGroundlayerAtmosphericeffectsTelescopesWavefrontsensorsWindprofilingTelemetry
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 tries to establish that wind profiling from the spatio-temporal cross-correlations of open-loop wavefront sensor telemetry can identify atmospheric turbulent layer velocities, and that on Maunakea the detected ground-layer speeds and directions broadly track independent weather-station and meteorological-model measurements. It reports a ground layer in about 80% of the selected telemetry files, free-atmosphere layers in fewer than 10%, and injection-recovery experiments showing sensitivity to layers with $r_0$ between 0.1 and 0.25 m and speeds from 1 to 36 m/s. If this is right, routine GLAO telemetry becomes a source of real-time atmospheric profiling for adaptive optics design and operation, without any additional hardware.

What carries the argument

The central object is the three-dimensional spatio-temporal covariance of wavefront sensor slope maps: equation (1) for auto-covariance of one sensor and equation (2) for cross-covariance between two sensors. Under the frozen-flow hypothesis, a turbulent layer crossing the pupil appears as a covariance peak that moves linearly with time; tracking that peak gives the layer's horizontal velocity, while the peak's offset at zero time lag in the cross-covariance gives the layer's altitude. The pipeline subtracts static and tip-tilt contributions, removes a background covariance to expose moving peaks, detects peaks at a three-standard-deviation threshold, and clusters the resulting tracks with a k-means algorithm.

What would settle it

A direct test would compare the temporal evolution of a single covariance peak's shape: if the frozen-flow hypothesis holds, the peak should move without significant broadening or amplitude loss over the window. Quantitatively, injecting a simulated layer with known velocity into real telemetry while adding a controlled decorrelation timescale shorter than the window should shift the recovered velocity away from the injected value; if it does not, the method is insensitive to the frozen-flow assumption. An independent balloon sounding at a known altitude would also settle whether the recovered velocities are unbiased.

Watch

Extended reading notes

Core claim

The central claim is that the motion of atmospheric turbulence layers can be extracted from the same open-loop wavefront sensor telemetry that a GLAO system already records. By computing auto-covariances of each wavefront sensor's slope maps and cross-covariances between different sensors, moving covariance peaks trace turbulent layers as they cross the pupil, and the offset of a cross-covariance peak at zero time lag encodes the layer's altitude. On the selected 'imaka runs, this method recovers a ground layer in the large majority of files and free-atmosphere layers only occasionally, and the detected speed and direction distributions follow external meteorological measurements with some outliers. The paper also shows, through simulated-layer injection into real telemetry, that the method is sensitive to layers across the typical range of speeds measured at Maunakea but misses fast, faint layers, which limits its use for upper-atmosphere profiling.

Load-bearing premise

The load-bearing premise is that each turbulent layer is a frozen pattern that simply translates across the telescope pupil without evolving during the correlation window of up to about 5.5 seconds; if a layer decorrelates or changes shape within that time, the inferred peak positions and velocities are biased.

Editorial extensions

If this is right

  • If the method is correct, wavefront telemetry from existing GLAO systems becomes a wind profiler with no extra instrument, enabling routine atmospheric monitoring during normal observations.
  • Ground-layer speed and direction estimates can be compared directly with summit weather stations and meteorological model forecasts, supporting site characterization and telescope scheduling.
  • Automated layer detection could feed real-time GLAO control decisions, though the current sensitivity limits mean fast, faint free-atmosphere layers should not be relied upon.
  • The known biases of the background-subtraction window imply that layer detections should be cross-checked with at least two subtraction lengths before being interpreted as meteorologically meaningful.
  • The method transfers naturally to other multi-conjugate or multi-wavefront-sensor AO systems, provided they archive open-loop slopes at comparable cadence.

Reading between the lines

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

  • Extension the paper leaves implicit: the altitude assignment is currently binary (in the cross-covariance means ground layer, otherwise free atmosphere), so allowing off-center cross-covariance peaks would turn the same telemetry into a continuous altitude profile.
  • The frozen-flow assumption is not tested on this data; comparing the width of the same covariance peak over increasing time lags would give a direct decorrelation timescale and could calibrate any velocity bias.
  • Since closed-loop telemetry can often be reconstructed from open-loop slopes, this technique may be applicable to GLAO systems that archive only closed-loop data, greatly expanding the available dataset for site characterization.
  • The roughly 45-degree offset between detected ground-layer directions and weather-station directions hints at local dome or ground-layer complexity; correlating with more than one nearby anemometer could separate local winds from layer winds.
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

5 major / 6 minor

Summary. The paper presents a wind-profiling method for ground-layer adaptive optics (GLAO) based on spatio-temporal auto- and cross-covariances of open-loop wavefront sensor slopes from the 'imaka instrument on the UH 2.2-meter telescope. The authors describe pre-processing, correlation computation, background subtraction, automated peak extraction, and k-means clustering to identify turbulent layer velocities. They inject simulated layers into representative on-sky telemetry to characterize detection sensitivity across speeds of 1-36 m/s and r0 values of 0.1-0.5 m. They report that a ground layer is identified in 70-80% of the selected telemetry files, free-atmosphere layers in less than 10%, and that the detected velocity distributions broadly match CFHT weather station and GFS model winds, though with some outliers. The central claim is that the method can recover layer velocities and that the measured distributions follow external meteorological measurements.

Significance. If validated, this work would demonstrate a practical use of GLAO open-loop telemetry for atmospheric wind profiling, with potential applications in AO design, scheduling, and control. The injection-recovery experiments are a concrete, controlled test of the detection pipeline and provide useful sensitivity estimates. The use of open-loop telemetry avoids the complications of reconstructing turbulence from closed-loop data, and the comparison to independent CFHT and GFS data is a step toward external validation. However, the validation is currently weakened by outcome-based data selection, a systematic direction offset that is not quantified or explained, and the absence of statistical tests for the distribution comparisons. The strengths are the automated processing of hundreds of files and the explicit acknowledgment of several algorithmic limitations.

major comments (5)
  1. [Section 2.1] The three observing runs are selected "based on quality of open loop telemetry, features in the data, and clarity of correlation peaks." This directly biases the reported detection fractions (70-80% ground layer, <10% free atmosphere) and the velocity distributions toward cases where the method works well. The conclusion in Section 6 that a ground layer is identified in 80% of files is a statement about this pre-selected sample, not about the method's typical performance on all 'imaka telemetry. Please quantify the selection criteria objectively (e.g., data-quality flags, frame dropouts, WFS availability) and report detection rates on the full sample or explicitly restrict the claims to the selected subset.
  2. [Section 5.1, Figure 8] The paper states that "the two main peaks of the found distribution are around 45 degrees off from the CFHT distribution." A systematic offset of the dominant direction peaks is not "some outliers"; it is a discrepancy in the main signal. The following sentence, "This offset seems to be from the last 5 years, looking at historical trends of wind directions captured by the weather station," does not explain or resolve the offset. No circular statistics, confidence intervals, or significance tests are provided. The central claim in Section 6 that the detected distributions "follow external meteorological measurements with some outliers" is not supported when the main directional peaks are systematically shifted. Please quantify the offset with circular statistics (e.g., mean resultant length, circular confidence intervals), test whether the offset is significant, and investigate potential causes such as coordinate/sign conventions, WFS orientation, or temporal mismatches between telemetry and weather data.
  3. [Sections 3.5 and 4.1] The injection-recovery experiments use simulated layers that are generated under the same frozen-flow translation assumption that underlies the covariance method in Eq. (1). These experiments therefore validate the peak-extraction and clustering pipeline but do not test whether real atmospheric layers remain coherent over the correlation window of up to 1000 frames (about 5.5 seconds at 180 Hz). If real layers decorrelate significantly over this window, the extracted peak positions and velocities will be biased. The paper cites reference [4] for frozen-flow analysis but does not measure the decay of covariance peak amplitude as a function of lag in its own telemetry. I request an analysis of peak amplitude versus lag (or an estimate of decorrelation timescales) to justify the maximum lag used, or a clear statement that the method assumes frozen flow without on-sky validation.
  4. [Section 5.1, speed comparison] The paper reports "a tail of higher speed detections" and "an over representation of slower speeds" relative to the CFHT and GFS distributions, but no quantitative measures are given: there are no error bars on the histograms, no goodness-of-fit statistics, and no per-file velocity correlations. The qualitative statement in Section 6 that the distributions "follow external meteorological measurements with some outliers" is therefore not quantitatively supported, especially given the systematic direction offset. Please provide numerical comparison metrics, such as a two-dimensional Kolmogorov-Smirnov test, circular-linear correlation, or per-night mean differences with uncertainties, and adjust the conclusions accordingly.
  5. [Section 3.4] The detection algorithm assumes that covariance peaks start at the center of the correlation map, which is valid for auto-covariances but not for cross-covariances when turbulent layers are at altitude. The paper acknowledges that layers between about 100 m and 600 m appear off-center in the cross-covariances and are consequently misidentified as free-atmosphere layers. This affects the reported ground-layer versus free-atmosphere classification in Figures 6 and 7 and the 80%/<10% detection fractions. The limitation is stated in Section 3.4, but it should be explicitly carried into the conclusions and, ideally, quantified in terms of the range of altitudes and velocities for which the current algorithm is reliable.
minor comments (6)
  1. [Section 1] Typo: "insight into its applciation and uses" should be "insight into its application and uses."
  2. [Section 5.1, Figure 8 caption] The caption reads "Comparison the the CFHT weather tower detections." It should read "Comparison with the CFHT weather tower detections."
  3. [Section 5.1] The text contains an incomplete sentence fragment: "The subsection" followed by "The speed distribution...". Please revise for readability.
  4. [Section 3.2, Eq. (1)] The notation in Eq. (1) is confusing: the sum $\sum_{ij}$ and the overlap normalization $O(\delta_i, \delta_j)$ are not fully defined, and the indices in $s_{i+\delta_i,j+\delta_j}$ are not explicitly bounded. Please clarify the summation limits and the definition of $O(\delta_i, \delta_j)$.
  5. [Section 3.5] The simulation uses four guide stars, but the recovery analysis "only consider the first three wavefront sensors" without justification. Since the number of WFSs varies on sky (3-5), please explain why three is representative or how the number of WFSs affects the detection sensitivity.
  6. [Section 4.2] The stated detection rates ("70-80%" ground layer, "less than 10%" free atmosphere) are given without the exact counts or uncertainties. Please cite the precise numbers from Figure 6 and, if possible, provide a statistical uncertainty (e.g., binomial confidence interval).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the layer-velocity comparison to independent CFHT/GFS data is not constructed from the fitted inputs.

full rationale

The derivation chain is self-contained. The wind-profiling measurement is obtained by spatio-temporal cross-covariances of open-loop WFS slopes (Eqs. 1 and 2), and the resulting layer velocities are compared against independent CFHT weather-station and GFS model data (Section 5). No parameter appearing in the claim of agreement is fitted to the external wind data; the only tuned parameters (detection threshold of 3 sigma and background-subtraction window of 50/200 steps) are calibrated by injection-recovery experiments on simulated layers (Sections 3.5 and 4.1) and are not adjusted to make the on-sky detections match CFHT or GFS. The GL/FA classification rests on the stated operational assumption that ground layers appear in both auto- and cross-covariances (Section 3.4), which is an acknowledged detection criterion rather than a hidden re-use of the conclusion. The one self-citation (Ref. [2] describing the 'imaka instrument) is not load-bearing for the profiling result. The paper's own limitations (frozen-flow assumption, off-center peak misclassification, directional offset in Fig. 8) weaken the strength of the agreement claim but do not make the derivation circular.

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

No new physical entities are introduced. The central claim rests on the frozen-flow assumption, an algorithmic assumption about peak origin, and two tuned processing parameters (detection threshold and subtraction window length), plus the ad hoc speed cut.

free parameters (3)
  • Detection threshold = 3 standard deviations
    Chosen in Section 4.1.1 based on injection recovery experiments; affects which covariance peaks are counted as detections.
  • Background subtraction window length = 50, 200, and 400 time steps
    Section 3.3 and 4.1.1; shorter windows suppress slow peaks, longer windows suppress fast peaks. Results from different windows are combined with a 5 m/s cut.
  • Speed cut for combining window lengths = 5 m/s
    Section 4.2: 'A cut at 5m/s is taken to combine the results of the two methods.' This ad hoc choice affects the final reported speed/direction distributions.
assumptions (3)
  • domain assumption Frozen flow: each turbulent layer translates across the pupil without evolving over the correlation window
    Section 3.2, Eq. (1): the covariance peak is interpreted as a translated replica of the wavefront; if layers decorrelate over up to 5.5 seconds, the extracted velocities are biased.
  • ad hoc to paper Covariance peaks for all detected layers start at the center of the correlation map at t=0
    Section 3.4: the algorithm only searches for peaks starting near the center; the paper acknowledges this misidentifies ground layers between 100 and 600 m as free-atmosphere layers.
  • domain assumption Dome seeing appears as a stationary covariance peak that can be removed by background subtraction without removing moving layer peaks
    Section 3.3: the central peak is attributed to dome seeing and subtracted via a rolling average; if it is not stationary, the subtraction biases the remaining layer speeds.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Wavefront Profiling via correlation of GLAO open loop telemetry." pith.science (2026). https://pith.science/paper/2NCKIJGZ

@misc{pith2026260810786,
  author       = {Pith},
  title        = {Pith review of: Wavefront Profiling via correlation of GLAO open loop telemetry},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2NCKIJGZ}},
  note         = {Machine review of arXiv:2608.10786}
}
read the original abstract

Adaptive Optics (AO) used in ground based observatories can be strengthened in both design and algorithms by a more detailed understanding of the atmosphere they seek to correct. Nowhere is this more true than on Maunakea, where a clearer profile of the atmosphere informs AO system development from the small separations of Extreme AO (ExAO) to the wide field Ground Layer AO (GLAO). Employing telemetry obtained from the 'imaka GLAO demonstrator on the University of Hawaii 2.2-meter telescope, we apply a wind profiling method that identifies turbulent layer velocities through spatial-temporal cross correlations of multiple wavefront sensors (WFSs). We compare the derived layer velocities with nearby wind anemometer data and meteorological model predictions of the upper wind speeds and discuss similarities and differences. The strengths and limitations of this profiling method are evaluated through successful recovery of injected, simulated layers into real telemetry. We detail the profilers' results, including the percentage of data with viable estimates, on four characteristic 'imaka observing runs on open loop telemetry throughout both winter and summer targets. We report on how similar layers are to external measures, the confidence of these results, and the potential for future use of this technique on other multi conjugate AO systems.

Figures

Figures reproduced from arXiv: 2608.10786 by the authors.

Figure 1
Figure 1. CFHT weather station wind velocities for ob [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. (First row) Dome seeing dominates the raw auto-covariance matrix. (Second row) An upward traveling peak is [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. (First row) Auto-covariance shows two peaks, one moving upward quickly, and another moving to the left. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: The successful recovery of injected layers across varying turbulence strengths and layer speeds. We see variation [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The number of layers across our files chosen. A majority of files have a least one ground laye (GL) identified [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: The distribution of speeds and directions across all found layers. A shorter length of background subtraction of [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Comparison the the CFHT weather tower detections. Speeds comparison shows a wide spread in speeds, [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Comparison of detected free atmosphere peaks with two pressures closest in altitude to expected layer location. [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

9 extracted references · 7 canonical work pages

  1. [4]

    Proceedings of the Third AO4ELT Conference , year=

    Analysis of the frozen flow assumption using gems telemetry data , author=. Proceedings of the Third AO4ELT Conference , year=

    work page

  2. [1]

    Annual Review of Astronomy and Astrophysics , volume =

    Guyon, Olivier , title =. Annual Review of Astronomy and Astrophysics , volume =. 2018 , doi =

    work page 2018

  3. [2]

    2017 , month =

    Leland McInnes and John Healy and Steve Astels , title =. 2017 , month =. doi:10.21105/joss.00205 , url =

    doi:10.21105/joss.00205 2017

  4. [3]

    Adaptive Optics Systems V , volume=

    Imaka: a ground-layer adaptive optics system on Maunakea , author=. Adaptive Optics Systems V , volume=. 2016 , organization=

    work page 2016

  5. [5]

    Annual Review of Astronomy and Astrophysics , volume=

    Adaptive optics for astronomy , author=. Annual Review of Astronomy and Astrophysics , volume=. 2012 , publisher=

    work page 2012

  6. [6]

    2014 , eprint=

    Cn2 and wind profiler method to quantify the frozen flow decay using wide-field laser guide stars adaptive optics , author=. 2014 , eprint=

    work page 2014

  7. [7]

    Applied optics , volume=

    Atmospheric turbulence profiling with SLODAR using multiple adaptive optics wavefront sensors , author=. Applied optics , volume=. 2008 , publisher=

    work page 2008

  8. [8]

    Wilson, R. W. , title = ". Monthly Notices of the Royal Astronomical Society , volume =. 2002 , month =. doi:10.1046/j.1365-8711.2002.05847.x , url =

    arXiv 2002

Show all 9 references
  1. [9]

    European Southern Observatory Conference and Workshop Proceedings , year = 2002, series =

    Ground Conjugate Wide Field Adaptive Optics for the ELTs. European Southern Observatory Conference and Workshop Proceedings , year = 2002, series =

    2002

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.