{"id":"acc3bf03-2850-4dcb-bb5a-ed239f997c6a","arxiv_id":"2608.10786","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A wind-profiling method applied to 'imaka GLAO open-loop telemetry detects ground-layer turbulence in ~70-80% of files and free-atmosphere layers in <10%, with speeds broadly matching external meteorological data.","lead":"This paper tests a method for measuring wind speeds of atmospheric turbulence layers by correlating wavefront sensor data from the 'imaka ground-layer adaptive optics system on Maunakea. It reports that ground-level turbulence is detected in most telemetry files, but high-altitude layers are rarely detected, and compares the results to weather station and model winds.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed agreement with external winds is not established: the paper reports a systematic 45-degree direction offset in Figure 8 but neither quantifies nor explains it.","rationale":"The reader's weakest assumption was frozen-flow, and that is a legitimate physical caveat. I do not dispute it. However, the single most load-bearing failure point for the paper's central claim is internal: the reported comparison that is supposed to validate the method contains a large, unquantified direction offset. Figure 8's own description undermines Section 6's 'follow external meteorological measurements' conclusion more directly than a decorrelation-timescale concern would. The frozen-flow assumption can be probed and may be acceptable over short windows; the 45-degree offset is visible in the paper's own headline comparison and is left unresolved. The concrete test above would settle whether the offset is real and significant, or an artifact of coordinate conventions. Because the underlying correlation method and injection-recovery study are still useful, I would keep the reader's CONDITIONAL verdict, but make the quantitative offset analysis and a statistical comparison to external winds explicit conditions of acceptance.","tokens_in":8147,"tokens_out":5828,"duration_ms":61284,"concrete_test":"Reproduce the Section 5.1 comparison with matched timestamps: for each detected GL cluster, pair it with the temporally nearest CFHT anemometer reading as done for Figure 8. Compute the circular mean and 95% confidence interval of the angular difference (detected minus CFHT) for the dominant peaks, and apply a circular Kuiper or Watson-Williams test against zero offset. If the confidence interval excludes 0 degrees by more than about 10 degrees, or the test rejects at p < 0.05, the paper must either trace the offset to a sign or coordinate convention error and recompute all velocities, or soften the claim that the detections follow external measurements. Apply the same procedure to the FA directions against the 250mb and 300mb GFS wind directions, since Section 5.2 makes a similar but weaker claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing claim, stated in Section 6, is that the detected layer velocities 'follow external meteorological measurements with some outliers.' The evidence for this is the comparison in Section 5.1 and Figure 8. That figure is reported as showing 'two main peaks of the found distribution are around 45 degrees off from the CFHT distribution.' A systematic 45-degree offset in the dominant direction peaks is not an outlier; it is a discrepancy in the main signal. The paper provides no circular statistics, confidence intervals, or a test of whether the offset is significant. The 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. Section 5.1 also reports a tail of higher detected speeds and an over-representation of slow speeds relative to the external data, again without a goodness-of-fit measure. Consequently, even if the frozen-flow assumption held and peak extraction were perfect, the central comparison does not support the conclusion as stated. The offset could be a coordinate or sign convention error, an epoch mismatch between telemetry and weather data, or a genuine method bias; the paper does not distinguish these possibilities, each of which changes the conclusion.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","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.","tokens_in":8307,"tokens_out":6193,"duration_ms":60932,"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":[{"comment":"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.","section":"Section 2.1"},{"comment":"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.","section":"Section 5.1, Figure 8"},{"comment":"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.","section":"Sections 3.5 and 4.1"},{"comment":"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.","section":"Section 5.1, speed comparison"},{"comment":"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.","section":"Section 3.4"}],"minor_comments":[{"comment":"Typo: \"insight into its applciation and uses\" should be \"insight into its application and uses.\"","section":"Section 1"},{"comment":"The caption reads \"Comparison the the CFHT weather tower detections.\" It should read \"Comparison with the CFHT weather tower detections.\"","section":"Section 5.1, Figure 8 caption"},{"comment":"The text contains an incomplete sentence fragment: \"The subsection\" followed by \"The speed distribution...\". Please revise for readability.","section":"Section 5.1"},{"comment":"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 $O(\\delta_i, \\delta_j)$.","section":"Section 3.2, Eq. (1)"},{"comment":"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.","section":"Section 3.5"},{"comment":"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).","section":"Section 4.2"}],"recommendation":"major_revision","confidential_remarks":"The topic is appropriate for astro-ph.IM and the injection-recovery experiments are a useful contribution. The main weaknesses are the outcome-based data selection in Section 2.1 and the lack of quantitative validation for the external comparisons, particularly the unexplained 45-degree direction offset in Figure 8. These issues are addressable with additional analysis, so a major revision is appropriate rather than rejection. I would also encourage the authors to check the coordinate conventions for the wind direction early in the revision process, as a sign error or rotation could explain the offset."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does something genuinely useful: it takes the established SLODAR-style correlation approach and applies it to a large set of 'imaka GLAO open-loop telemetry, with an automated peak-extraction pipeline and an injection-recovery sensitivity study. The injection experiments are a real strength—they give honest limits on what the method can detect (r0 between 0.1 and 0.25 m, speeds 1–36 m/s) and they are used to set detection thresholds, which is legitimate calibration rather than circularity. The authors also clearly acknowledge the frozen-flow assumption and the 100–600 m misidentification issue. That is more honest than many AO telemetry papers.\n\nThe problem is the central comparison to external data. The conclusion in Section 6 that the detected distributions \"follow external meteorological measurements\" is not supported by the evidence in Figure 8. The two main peaks in the detected ground-layer directions are about 45 degrees off from the CFHT distribution. That is not an outlier; it is a systematic discrepancy in the dominant signal. The paper's explanation—\"This offset seems to be from the last 5 years\"—is vague and does not resolve whether it is a coordinate convention error, a sign flip, an epoch mismatch, or a genuine method bias. There are no circular statistics, confidence intervals, or goodness-of-fit measures anywhere in the comparison. The same is true for the speed distributions, where a tail of faster detections and an over-representation of slow speeds are reported without any quantitative assessment.\n\nThe selection bias is real but secondary: the telemetry files were chosen for \"quality and clarity of correlation peaks,\" which inflates the 70–80% ground-layer detection fraction. The authors do not apply the pipeline to all available files to estimate what an unbiased detection rate would be. No code or data are provided, which makes the injection-recovery study hard to reproduce independently.\n\nStill, the method is sound and the application is new. The paper is worth refereeing seriously, but it needs major revisions before publication: quantify the agreement with external data, resolve or explain the 45-degree offset, run the detector on a non-preselected sample, and release at least the telemetry subsets or code. I would not cite it in its current form, but I would revisit it after those fixes.","headline":"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.","tokens_in":8909,"tokens_out":1204,"would_cite":false,"duration_ms":13266,"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":"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.","keywords":["Adaptive Optics","Optical turbulence","Ground layer adaptive optics","Atmospheric effects","Telescopes","Wavefront sensors","Wind profiling","Telemetry"],"falsifier":"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.","tokens_in":7886,"feed_emoji":"🌬️","tokens_out":5168,"duration_ms":48971,"temperature":0.7,"pith_summary":"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.","feed_headline":"Telemetry wind layers match Maunakea weather data in 80% of files","feed_subtitle":"Ground-layer speeds from 'imaka wavefront-sensor correlations broadly agree with weather station and forecast winds.","key_machinery":"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.","core_discovery":"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.","pith_inferences":["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."],"forward_implications":["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."],"supporting_citations":[{"why":"Supplies the spatio-temporal correlation formalism (SLODAR) for measuring turbulence altitude with Shack-Hartmann wavefront sensors, which this paper adapts to GLAO telemetry.","marker":"[3]"},{"why":"Analyzes the frozen-flow assumption with GEMS telemetry, the premise on which the covariance-peak interpretation rests.","marker":"[4]"},{"why":"Extends SLODAR wind profiling to multiple wavefront sensors, the multi-WFS approach this paper applies to GLAO.","marker":"[5]"},{"why":"Provides a method to quantify frozen-flow decay and is used in the paper to attribute the stationary central covariance peak to dome seeing.","marker":"[6]"},{"why":"Describes the 'imaka GLAO system and its wavefront sensor telemetry, which forms the dataset analyzed here.","marker":"[2]"}],"fun_headline_variants":["GLAO telemetry reveals wind layers matching Maunakea forecasts","Wind profiling from GLAO telemetry matches weather stations","Wavefront sensor correlations map Maunakea wind layers","GLAO telemetry yields wind speeds close to weather data","Correlating GLAO sensors gives wind profiles matching forecasts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"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.","fun_headline_variants_meta":{"raw":{"variants":["GLAO telemetry reveals wind layers matching Maunakea forecasts","Wind profiling from GLAO telemetry matches weather stations","Wavefront sensor correlations map Maunakea wind layers","GLAO telemetry yields wind speeds close to weather data","Correlating GLAO sensors gives wind profiles matching forecasts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00139,"raw_usage":{"total_tokens":5630,"prompt_tokens":954,"completion_tokens":4676,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":570,"completion_tokens_details":{"reasoning_tokens":4590}},"tokens_in":570,"tokens_out":4676,"duration_ms":30130,"temperature":1.0,"reasoning_tokens":4590,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:22:23.159602+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"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.","supporting_citations":[{"cited_title":"Adaptive Optics Systems V , volume=","cited_arxiv_id":null,"evidence_quote":"Supplies the spatio-temporal correlation formalism (SLODAR) for measuring turbulence altitude with Shack-Hartmann wavefront sensors, which this paper adapts to GLAO telemetry."},{"cited_title":"Proceedings of the Third AO4ELT Conference , year=","cited_arxiv_id":null,"evidence_quote":"Analyzes the frozen-flow assumption with GEMS telemetry, the premise on which the covariance-peak interpretation rests."},{"cited_title":"Annual Review of Astronomy and Astrophysics , volume=","cited_arxiv_id":null,"evidence_quote":"Extends SLODAR wind profiling to multiple wavefront sensors, the multi-WFS approach this paper applies to GLAO."},{"cited_title":"2014 , eprint=","cited_arxiv_id":null,"evidence_quote":"Provides a method to quantify frozen-flow decay and is used in the paper to attribute the stationary central covariance peak to dome seeing."}],"review_version":1}