{"id":"9c4187ab-73bb-4a5c-87d7-6076906f591d","arxiv_id":"2412.00231","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Soaring golden eagles show extreme upward acceleration bursts during turbulent gusts, with a nonlinear response that suggests they may harvest energy from turbulence.","lead":"Golden eagles in soaring flight experienced sharp upward acceleration bursts during strong turbulent gusts, sometimes more than three times gravity. The authors argue these birds may harvest energy from turbulence rather than fight it, a finding that could reshape how we think about animal flight and small drone design.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline ratcheting/gust-harvesting claim depends on attributing extreme accelerations to external gusts; the accelerometer alone cannot exclude wing adjustments, body tucks, or harness/sensor motion, and the paper's own 'unknown origin' oscillation admits this.","rationale":"The paper is a careful field-data analysis: long-duration accelerometer records from free-flying eagles, a physically motivated comparison with wind-tunnel turbulence, and a clean demonstration that golden-eagle gliding acceleration tails are broader and more asymmetric than a linear response would predict. The flatness crossover at τ≈0.4 s matching the model's predicted τc≈0.3 s is a genuinely interesting prediction, though k is estimated from the same data and σwz is not stated. The central claim, however, is causal: extreme upward accelerations are 'interactions with gusts' and evidence of a 'ratcheting mechanism' that harvests turbulent energy. For that claim, the data must show that the extreme accelerations are not produced by the eagle's own wing/body adjustments or by the harness. The paper's own statements—an oscillation 'whose origin is unknown to us' that 'could be related to the way the sensors were attached' and the Discussion's explicit assumption that all extreme events are gust interactions—show that this identification is not secured. The classification into flapping versus non-flapping removes only periodic flapping; aperiodic wing tucks, tail movements, and postural changes are documented responses of soaring raptors to gusts (Reynolds et al. 2014; Cheney et al. 2020, cited by the authors) and would appear in the gliding sample. Since no independent wind measurement is co-located with the bird, the conditional statistics cannot distinguish external forcing from the bird's own dynamics. Thus the strongest finding is conditional: the asymmetry is real, but its gust-harvesting interpretation needs a control that separates body-generated accelerations from aerodynamic forcing. This is exactly the reader's weakest assumption, so I agree with the reader's CONDITIONAL verdict and recommend no change.","tokens_in":13633,"tokens_out":6410,"duration_ms":64050,"concrete_test":"Recompute the conditional averages and flatness (Figs. 2b and 3) using only extreme events for which the body-fixed x- and y-acceleration channels remain within their quiet-gliding baseline (no detectable pitch/roll change or wing tuck) and for which the short-timescale oscillation flagged as being of unknown origin is absent in a band-pass (e.g., 5–20 Hz) around the event. If the H=±5 upward asymmetry and the τ<0.4 s flatness divergence persist in this subset, the gust attribution is supported; if they weaken or disappear, active body/harness motion is the source, and the gust-harvesting claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the extreme upward accelerations recorded during gliding are produced by atmospheric wind gusts acting on the bird, not by the eagle's own active or passive mechanical responses or by the instrument package. The paper's measurement is a body-fixed accelerometer; there is no independent, co-located measurement of the local wind. The analysis distinguishes only flapping from non-flapping flight (Materials and Methods, 'Classification of flight behaviors'), so partial flaps, wing tucks, tail or angle-of-attack adjustments, body accelerations from the bird's control responses, and harness resonance all remain in the 'gliding' class. The authors explicitly flag an unresolved contaminant: 'We note an oscillation with a short timescale ... whose origin is unknown to us' and say it 'could be related to the way the sensors were attached to the eagles or to wing adjustments made in response to turbulent gusts.' In the Discussion they then write, 'Assuming that all extreme events observed in the interval 0.2 s < τ < 2 s are due to interactions with gusts, which is what the statistics suggest.' That assumption is circular for this question: the statistics are statistics of the eagles' measured accelerations, so they cannot by themselves identify the aerodynamic cause. If a substantial fraction of the H=±5 events are body-generated (e.g., the wing tucks documented by Reynolds et al. 2014, or harness vibration), the upward bias and the τ−1 flatness rise that the nonlinear model attributes to gust amplification could instead be behavioral or mechanical, and the 'first quantitative evidence in favor of turbulent gust harvesting by wildlife' would not be established. The empirical asymmetry is interesting, but the gust attribution is the load-bearing step that is currently unsupported by an independent control.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes 40 Hz accelerometer data from six golden eagles and six bald eagles in free flight. It reports that vertical acceleration differences in soaring golden eagles have heavy-tailed distributions, with excursions up to 25 standard deviations, that the conditional averages around extreme events resemble turbulent vortex passages, and that extreme upward acceleration events break the up/down symmetry of the underlying wind statistics. The authors propose a nonlinear response model az = (wz/τb)(1 + k wz τb/ℓ), estimate k ≈ 0.6 from the up/down asymmetry of extreme events, and claim the model predicts a crossover timescale τc ≈ 0.3 s near which acceleration flatness rises as τ^{-1}, consistent with observed small-scale intermittency. They interpret the results as a 'ratcheting mechanism' and 'the first quantitative evidence in favor of turbulent gust harvesting by wildlife.'","tokens_in":13936,"tokens_out":6416,"duration_ms":58385,"significance":"The statistical work is careful: the flapping classification uses cross-correlation with template flaps, the golden-eagle versus bald-eagle comparison separates flapping-related intermittency from turbulence-like intermittency, and sensitivity checks on thresholds are reported. If the extreme accelerations are indeed caused by atmospheric gusts, the observed upward asymmetry and small-scale intermittency enhancement would be a novel and physically interesting example of nonlinear interaction between a flying animal and boundary-layer turbulence, with implications for the design of small aerial vehicles and for interpreting bio-logging data. However, the paper's headline conclusion—gust harvesting—requires an energy budget and a direct attribution of accelerations to wind, neither of which is provided; the currently available evidence supports a more modest claim about nonlinear, asymmetric acceleration statistics during soaring.","major_comments":[{"comment":"The attribution of extreme accelerations to external gusts is not established. The measurements are body-fixed accelerations; there is no independent, co-located wind measurement, and the flight-behavior classification separates only flapping from non-flapping, leaving partial flaps, wing tucks, tail adjustments, and harness/sensor dynamics inside the gliding class. The paper itself acknowledges an oscillation 'whose origin is unknown to us' that 'could be related to the way the sensors were attached ... or to wing adjustments,' and in the Discussion assumes 'that all extreme events observed in the interval 0.2 s < τ < 2 s are due to interactions with gusts, which is what the statistics suggest.' That sentence is circular for the causal question: the statistics are statistics of the eagles' measured accelerations and cannot by themselves identify the aerodynamic cause. Because the gust-harvesting claim depends on this assumption, the manuscript needs either a direct test (e.g., temporal or spatial correlation with independently measured atmospheric turbulence, or a kinematic control using the lateral and longitudinal channels to detect wing adjustments) or a substantial reframing of the conclusions as conditional on the gust assumption.","section":"Vortical structure of the eagles' acceleration / Discussion and conclusions"},{"comment":"The nonlinear coefficient k is estimated from the asymmetry of the very extreme events (the relation a+z/a−z ≈ −1 + k(τb^2/ℓ)(a+z − a−z)), and the same k is then used to derive the flatness enhancement Fδaz ≈ (1 + 4kσwzτb/ℓ)Fδwz and the crossover timescale τc = 1/(4V kσwzτb). Consequently, the agreement of the predicted crossover near 0.3 s with the observed crossover near 0.4 s is an internal consistency check, not an independent prediction of the model. The abstract's phrase 'predicts the scale at which symmetry breaks' is thus overstated. To make the claim non-circular, the authors would need to estimate k from one subset of events (or from a different observable) and then test the flatness and crossover prediction on independent data.","section":"Nonlinear model of the eagles' response to gusts"},{"comment":"The model and the flatness comparison are constructed asymmetrically: the text states that the nonlinearity is not physical for large negative wz, so the flatness Fδaz is calculated only for positive excursions in wz, while the turbulent wind flatness Fδwz is computed from both signs. The derived scaling Fδaz ∼ (1 + 4kσwzτb/ℓ)Fδwz therefore depends on this sign restriction and on the assumptions of local isotropy and ⟨δwΣw⟩ = 0. These restrictions should be stated as part of the model definition, and the comparison should be justified or the metric defined consistently; otherwise the predicted 'stronger intermittency than turbulence' may be an artifact of the asymmetric truncation.","section":"Nonlinear model of the eagles' response to gusts"},{"comment":"The claim of gust harvesting requires an energy budget. A ratcheting mechanism that increases the mean mechanical energy of the eagle must show that the work done by the fluctuating wind on the bird is positive on average after accounting for induced drag, the cost of the bird's control responses, and the vertical component of the flight path. The paper states only that the eagles experienced upward accelerations and that these are 'in the eagles' interest to stay aloft'; it does not show that the eagles extract net energy from the gusts. The phrase 'first quantitative evidence in favor of turbulent gust harvesting' is therefore not supported by the presented analysis. At minimum, the manuscript should identify the energy (or power) balance needed and state why the observed asymmetry implies net positive energy gain, or soften the conclusion to 'consistent with' rather than 'evidence in favor of.'","section":"Discussion and conclusions"},{"comment":"The estimate that about 20% of flight time was spent in extreme events with gust ratio GR > 0.3 uses the linear relation |δaz| = |δwz|/τb to convert acceleration tails to wind velocity tails, but this is the same regime for which the paper argues linearity breaks down. Because the nonlinear model amplifies upward accelerations, the mapping from δaz to δwz is not one-to-one, and the 20% figure is model-dependent. The authors should provide the estimate under the linear model and under the nonlinear model, or explicitly label the number as a linear-model estimate whose uncertainty includes the nonlinearity.","section":"Discussion and conclusions"}],"minor_comments":[{"comment":"In the flap detection description, 'we labeled the time corresponding to the peak of mean flap as T = 0' should be 't = 0' to avoid confusion with the averaging time T in the flap period notation.","section":"Materials and Methods, Classification of flight behaviors"},{"comment":"In the sentence 'we selected δ(t) within ±5T', the symbol δ(t) should be written as δaz(t, τ) (or δa(t, τ)) to be consistent with the definitions used elsewhere in the paper.","section":"Materials and Methods, Conditional Averaging"},{"comment":"The caption states that the findings hold for any τ between 0.2 and 4 s, but the figure displays only τ = 0.2 s and τ = 2 s; the caption should either show the additional timescales or qualify the statement as applying to the tested range.","section":"Fig. 2 caption"},{"comment":"Because the central claims are driven by rare tail events (N ≈ 10^3 for H = ±5), the data availability statement should clarify how an independent researcher can obtain the full dataset and the analysis scripts, since the current 'available upon request with permission' policy limits verification of the extreme-event statistics.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":"This is a high-profile claim from a group with relevant prior work (Laurent et al. 2021). The manuscript's own caveats are more candid than is typical, but the gap between the data and the 'gust harvesting' conclusion is large. I think the paper can be made acceptable with major revision that either adds direct gust attribution or reframes the conclusion as conditional on the gust assumption; I would not accept it in its current form. The lack of a co-located wind measurement and the acknowledged 'unknown origin' oscillation are the load-bearing weaknesses."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First: this is a carefully analyzed field dataset, and the main empirical finding — that soaring golden eagles show upward-biased extreme acceleration bursts that are more intermittent than turbulence on short timescales — looks real. It deserves to be published in some form. The comparison with bald eagles as flapping controls is a sensible way to separate behavioral from aerodynamic contributions, and the conditional averages around extreme events are a nice use of turbulence tools.\n\nWhat is actually new: the nonlinear response model with a quadratic correction to the linear relation between vertical wind and acceleration, and the empirical symmetry breaking at H=±5. The crossover timescale prediction near 0.3 s matching the observed 0.4 s is suggestive, though not a free prediction because k is fitted from the same data.\n\nThe soft spots are the usual ones for animal-borne accelerometer studies, but they bite here because the headline claim depends on them. You have no independent measurement of the local wind at the bird. The sensor is on the back; it records the bird's body acceleration, not the wind. The paper itself flags an oscillation of unknown origin that could be harness resonance or wing adjustments. The Discussion then assumes all extreme events between 0.2 s and 2 s are gust interactions, 'which is what the statistics suggest' — but the statistics are of the bird's accelerations, so that reasoning is circular for the attribution question. Wing tucks, partial flaps, or tail adjustments could produce upward-biased acceleration spikes without any gust harvesting. The model's k is fitted to the asymmetry it is supposed to explain, so the flatness prediction is not independent.\n\nThat said, the paper is honest about its limits and does not overclaim in the body; the abstract's 'first quantitative evidence' is stronger than what the data support. The data are proprietary, so independent checks will be hard.\n\nMy take: this is a solid empirical contribution that would be stronger as a careful description of asymmetric extreme accelerations in soaring eagles, with the harvesting interpretation clearly flagged as a hypothesis. A good referee could push the authors to either get co-located wind data or explicitly model the bird's own motion. That is a real revision, not a desk reject. Send it to peer review.","headline":"Careful statistics show a real upward asymmetry in eagles' extreme accelerations, but the gust-harvesting claim outruns what an accelerometer alone can prove.","tokens_in":14498,"tokens_out":2424,"would_cite":false,"duration_ms":22697,"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":"Golden eagles gain lift from violent turbulent gusts","keywords":["eagle flight","turbulent gusts","gust harvesting","intermittency","nonlinear aerodynamics","soaring","accelerometry","lift amplification"],"falsifier":"An experiment that would settle it: mount the same accelerometer package on a rigid glider or a taxidermied bird in a wind tunnel with intermittent gusts; if the upward bias disappears, the eagle's active body movements are the cause rather than the gust.","tokens_in":13426,"feed_emoji":"🦅","tokens_out":7421,"duration_ms":61903,"temperature":0.7,"pith_summary":"This paper claims that soaring golden eagles do not simply endure turbulence; they extract upward momentum from short, intense gust bursts. Accelerometer data from wild eagles show vertical accelerations up to 25 standard deviations from the mean, more than three times gravity, with upward bursts more frequent and stronger than downward ones. Such asymmetry is impossible under linear gust-response models, so the authors introduce a nonlinear correction in which strong gusts produce extra lift. They argue this constitutes the first quantitative evidence of turbulent gust harvesting by wildlife, meaning turbulence is an energy source rather than only a disturbance for animals and potentially for small aircraft.","feed_headline":"Golden eagles gain lift from violent turbulent gusts","feed_subtitle":"Accelerometer data show upward bursts beyond any linear model — first evidence of gust harvesting in wildlife.","key_machinery":"The central machinery is a conditional-averaging statistic from turbulence research: average the acceleration traces around local extrema of the acceleration difference $\\delta a_z(t,\\tau)$ to reconstruct the signature of a vortex encounter. The load-bearing identity is the nonlinear gust-amplification model $a_z = (w_z/\\tau_b)(1 + k w_z \\tau_b/\\ell)$, with $k>0$, which ties the asymmetry between upward and downward excursions to a predicted crossover in the flatness $F_{\\delta a_z}$ at $\\tau_c \\approx 0.3$ s. The flatness, defined as the normalized fourth moment of a distribution, measures tail breadth and is the key observable that separates turbulence-like scaling from the eagle-specific intermittency.","core_discovery":"On timescales below about 2 s, the vertical acceleration differences of soaring golden eagles are far more intermittent than turbulence itself and are biased upward. Conditional averages of extreme events (|H| ≥ 5 standard deviations) show that the eagles are pushed up more strongly than down, breaking the symmetry found in turbulent wind statistics at small scales. The authors model this with a single-parameter nonlinear law, $a_z = (w_z/\\tau_b)(1 + k w_z \\tau_b/\\ell)$, with $k \\approx 0.6 \\pm 0.1 > 0$ indicating gust amplification rather than stall or mitigation. The model predicts a crossover near $\\tau_c \\approx 0.3$ s between turbulence-like intermittency (flatness $\\sim \\tau^{-0.2}$) and a steeper $\\tau^{-1}$ rise, matching the data. The conclusion is a ratcheting mechanism: eagles repeatedly convert strong upward gust impulses into lift, and about 20% of soaring flight time is spent in gusts strong enough to trigger nonlinear aerodynamic changes.","pith_inferences":["If the ratcheting picture holds, the energy gained from gusts could partially offset the metabolic cost of flapping and gliding, a net energy budget that the paper does not attempt to calculate.","An extension of the model suggests that smaller birds with shorter wing chords should show an earlier crossover and stronger short-scale intermittency, a testable prediction across species.","The coefficient k could be estimated per individual from the flatness data; if k varies with wind conditions or body condition, it would point toward active control rather than purely passive aerodynamics.","A direct experimental test would be to expose a fixed-wing glider or a non-flapping model in a wind tunnel to intermittent gusts and check whether the upward bias appears without active wing motion; if it does, the effect is passive aerodynamic amplification."],"forward_implications":["If the ratcheting mechanism is correct, turbulence must be treated as an energy input in the flight energetics of soaring birds, not merely as a dissipative disturbance.","The measured 20% of soaring time spent in strong gusts implies that unsteady, nonlinear aerodynamics is the norm in boundary-layer soaring flight, not a rare edge case.","The predicted crossover near 0.3 s provides a target for when gust exploitation outweighs linear response, a scale directly relevant to the design of small aircraft and drones.","The same conditional-averaging and flatness analysis could be applied to accelerometer data from other volant species to detect gust harvesting wherever it occurs."],"supporting_citations":[{"why":"Supplies the linear model and the timescale over which turbulence dominates eagle accelerations, the baseline this paper extends.","marker":"[5]"},{"why":"Provides the conditional velocity-difference averaging method used to extract vortex-like structure from extreme events.","marker":"[47]"},{"why":"Gives the standard phenomenology of small-scale turbulence intermittency and flatness scaling used as the turbulence reference.","marker":"[13]"},{"why":"Foundational structure-function relations for turbulence, used in defining velocity increments and the isotropy assumptions.","marker":"[49]"},{"why":"Documents wing tucks in response to turbulence in soaring eagles, the competing behavioral explanation the analysis must distinguish from.","marker":"[28]"},{"why":"Provides the ~tau^-0.2 flatness scaling of transverse velocity increments in atmospheric turbulence, the curve the eagle data are compared with.","marker":"[56]"},{"why":"Frozen-flow hypothesis linking temporal and spatial statistics, used to map the eagle acceleration timescales to vortex length scales.","marker":"[54]"}],"fun_headline_variants":["Eagles harvest turbulent gusts for lift: first proof","Gust harvesting: eagles turn turbulence into lift","Violent gusts lift eagles: first evidence of harvesting","Ratcheting turbulence: eagles gain lift from gust bursts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the extreme vertical accelerations recorded during gliding are caused by atmospheric wind gusts rather than by the eagles' own wing adjustments, partial flaps, or sensor motion.","fun_headline_variants_meta":{"raw":{"variants":["Eagles harvest turbulent gusts for lift: first proof","Gust harvesting: eagles turn turbulence into lift","Violent gusts lift eagles: first evidence of harvesting","Ratcheting turbulence: eagles gain lift from gust bursts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000917,"raw_usage":{"total_tokens":3976,"prompt_tokens":1023,"completion_tokens":2953,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":639,"completion_tokens_details":{"reasoning_tokens":2890}},"tokens_in":639,"tokens_out":2953,"duration_ms":21913,"temperature":1.0,"reasoning_tokens":2890,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:35:41.660458+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An experiment that would settle it: mount the same accelerometer package on a rigid glider or a taxidermied bird in a wind tunnel with intermittent gusts; if the upward bias disappears, the eagle's active body movements are the cause rather than the gust.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the linear model and the timescale over which turbulence dominates eagle accelerations, the baseline this paper extends."},{"cited_title":"& Kawashima, Y","cited_arxiv_id":null,"evidence_quote":"Provides the conditional velocity-difference averaging method used to extract vortex-like structure from extreme events."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the standard phenomenology of small-scale turbulence intermittency and flatness scaling used as the turbulence reference."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Foundational structure-function relations for turbulence, used in defining velocity increments and the isotropy assumptions."},{"cited_title":"V., Thomas, A","cited_arxiv_id":null,"evidence_quote":"Documents wing tucks in response to turbulence in soaring eagles, the competing behavioral explanation the analysis must distinguish from."},{"cited_title":"& Sreenivasan, K","cited_arxiv_id":null,"evidence_quote":"Provides the ~tau^-0.2 flatness scaling of transverse velocity increments in atmospheric turbulence, the curve the eagle data are compared with."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Frozen-flow hypothesis linking temporal and spatial statistics, used to map the eagle acceleration timescales to vortex length scales."}],"review_version":1}