{"id":"5f54360c-032a-4d28-9a29-f2ebd6be7aa0","arxiv_id":"2412.03384","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Using coarse-grained Boussinesq simulations, strong vertical drafts are associated with enhanced downscale kinetic transfer and bidirectional potential-energy transfer near the buoyancy scale.","lead":"This paper refines a space-filtering (coarse-graining) method and applies it to simulations of stratified turbulence, showing that regions with strong vertical velocity drafts carry enhanced and sometimes bidirectional energy transfers across scales. It matters because it offers a local view of how energy moves in intermittent ocean and atmosphere flows, where dissipation is concentrated in small regions.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Draft-conditioned flux statistics rest on unvalidated pointwise coarse-graining: only volume-averaged fluxes are benchmarked, and the 1D Butterworth filter admits negative local subscale kinetic energy.","rationale":"The reader's weakest assumption is the right one. The derivation of the filtered Boussinesq energy equations is standard, and the volume-average comparison in Sec. IV A is genuine evidence that the implementation computes global fluxes correctly. But global correctness does not certify pointwise or subdomain accuracy: a local filter can satisfy the volume-averaged equality while pointwise S fields are distorted, especially with a non-positive kernel such as Butterworth; the admitted tr[T^uu] < 0 is a concrete symptom. The paper itself also warns, citing Ref. [3], that partial fluxes may not capture the actual energy cascade rate in geophysical flows. Since every result in Sec. V is a conditional average over w-bins of these local S fields, filter-dependent local leakage is a direct threat to the abstract's central claim. A causal-language concern also exists — conditional correlation alone does not establish that drafts trigger transfers — but that concern is secondary and rephrasing-level, whereas the measurement premise would change the content if wrong. The proposed filter-substitution test is feasible on the same stored snapshots and would settle the concern. Thus the CONDITIONAL verdict stands unchanged.","tokens_in":19795,"tokens_out":7154,"duration_ms":82666,"concrete_test":"Recompute the Sec. V conditional fluxes for Run II using a sharp spectral low-pass filter and, separately, a top-hat or Gaussian kernel at the same cutoffs k⊥ = 7 and k∥ = 7, with identical bins and snapshots. Compare the high-|w| profiles of ⟨Su⟩, ⟨Sθ⟩, and ⟨Stot⟩ across filters. If the enhanced forward transfer, the ⟨Sθ⟩ zero-crossing near k⊥ ≈ 15, and the parallel upscale/downscale split around k∥ ≈ kB survive quantitatively, the concern is discharged; if sign inversions shift by more than the filter width or disappear, the local-flux interpretation is not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim — that high-|w| draft regions show enhanced forward kinetic transfer, bidirectional potential transfer around k⊥≈15, and an upscale/downscale split near k∥≈kB — is built from conditional averages of pointwise subfilter fluxes in Sec. V. What is validated in Sec. IV A is the volume average of the same terms against Fourier fluxes, and only on Run I (kF = 20, Re ≈ 97), not the pointwise or subdomain values used for Run II in Sec. V. A pointwise flux estimate requires that the local filtered stress and heat-flux terms are meaningful at every point. The authors explicitly note that the Butterworth filter can produce tr[T^uu] < 0 pointwise (Sec. III), so the local subscale kinetic energy is not sign-definite. With the one-dimensional kernels G(4)(k⊥) and G(4)(k∥), scales in the un filtered directions are not removed, so local signed fluxes can mix horizontal and vertical scale content in a way that global plane-integrated comparisons cannot detect. If this bias is stronger in regions of large |w|, the reported amplification factors and sign inversions could be artifacts even though the global comparison is excellent. This is a measurement-validity premise, not a challenge to the algebraic derivation of the filtered equations.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper adapts the coarse-graining (space-filtering) method to the Boussinesq equations for stably stratified turbulence, deriving filtered kinetic, potential, and total energy equations with sub-filter-scale transfer terms and a kinetic-to-potential conversion term. The authors implement one-dimensional Butterworth filters to define parallel and perpendicular reduced fluxes, and validate the volume-averaged terms against classical Fourier fluxes for a moderately forced run. They then apply the method to a strongly intermittent run (Run II, Fr ≈ 0.08) and compute conditional averages of the local sub-filter fluxes in bins of the standardized vertical velocity |w/σ_w|. The central claim is that in regions of strong vertical drafts the kinetic and total energy fluxes are enhanced and forward at intermediate perpendicular scales, the potential energy flux becomes bidirectional around k⊥ ≈ 2k_B, and along the parallel direction the total flux is upscale for k∥ < k_B and downscale for k∥ > k_B. The paper also reports a temporal correlation between the vertical-velocity kurtosis and the volume-averaged flux modulation.","tokens_in":20051,"tokens_out":13482,"duration_ms":143413,"significance":"If the conditional results are robust, the paper provides a valuable local diagnostic for stratified turbulence and extends previous work on large-scale intermittency by connecting extreme vertical drafts to scale-to-scale energy transfer and kinetic-potential energy exchange. The derivation of the filtered Boussinesq equations is transparent and the volume-averaged comparison with Fourier fluxes is a genuine strength, as is the use of conditional statistics over many samples. The qualitative prediction that high-|w| regions host enhanced and bidirectional transfers is falsifiable and of interest for parameterizations of mixing and dissipation in geophysical flows. However, the causal wording of the abstract and conclusions goes beyond the correlational evidence, and the local validity of the pointwise flux estimates needs additional support before the central claim can be accepted as stated.","major_comments":[{"comment":"The main physical claim is built from conditional averages of pointwise sub-filter fluxes in Sec. V (Figs. 5 and 6), but the only quantitative benchmark in Sec. IV A (Fig. 2) validates volume-averaged terms, and only for Run I. The authors themselves note in Sec. III that the Butterworth filter can give pointwise tr[T^uu] < 0, and with one-dimensional filters the local flux at a point contains contributions from all scales in the unfiltered directions. It is therefore not possible to exclude a priori that filter artifacts are stronger in regions of large |w|, which could bias the reported amplification factors and sign inversions even though the global comparison is excellent. I ask for robustness tests directly on the quantities used in Sec. V: for example, recompute the draft-conditioned profiles with a different filter order or a different cutoff, check sensitivity to the binning in |w/σ_w|, and, if feasible, compare localized fluxes with a sharp-filter or local spectral estimate. This is a measurement-validity issue that is load-bearing for the central claim.","section":"Sec. V and Sec. IV A"},{"comment":"The abstract states that vertical drafts 'are indeed able to trigger upscale and downscale energy transfers', and Sec. VI says that drafts 'act as a mechanism for locally converting energy from kinetic to potential'. The evidence presented, however, consists of conditional averages over snapshots and a temporal correlation between the kurtosis K_w and volume-averaged fluxes (Fig. 7). These statistics establish association, not causation: the enhanced fluxes could be produced by the same dynamics that create the drafts, or by pre-existing large-scale strain, rather than by the drafts themselves. Please either soften the language to association ('are associated with', 'co-localize with') or provide a causal analysis such as time-lagged conditional fluxes or tracking of individual draft events. As written, the central claim overstates what the data can support.","section":"Abstract and Sec. VI"},{"comment":"Equations (17) and (18) define the Fourier fluxes Π_u and Π_θ from the nonlinear advective transfer only. The 'conservative fluxes' Φ_u = S_u + N⟨θ̃w̃⟩ and Φ_θ = S_θ − N⟨θ̃w̃⟩, shown as red lines in Fig. 2, additionally include the kinetic-to-potential conversion term. Comparing these two quantities as if they were the same object is not apples-to-apples; indeed the text itself notes that the conversion term dominates at k⊥ ≳ 70, where the comparison cannot be expected to hold. I recommend comparing ⟨S_u⟩ directly with Π_u and ⟨S_θ⟩ directly with Π_θ, and plotting the cumulative buoyancy conversion separately if it is to be used as a budget term. This is needed to make the validation of the individual energy channels unambiguous.","section":"Sec. IV A, Fig. 2"}],"minor_comments":[{"comment":"The first sentence of Sec. IV A refers to 'a stably stratified flow Run II' but then identifies the simulation as Run I in Table I (kF = 20); this is a typo that should be corrected.","section":"Sec. IV A"},{"comment":"The values ⟨ε_V⟩ ≈ 0.288 and ⟨ε_θ⟩ ≈ 0.024 are cited without a definition; please specify whether these are volume-averaged dissipation rates, transfer rates, or filtered energy budget terms.","section":"Sec. V A"},{"comment":"The column labels in Table I are ambiguous ('k F', 'n p'); the caption should define every column, including the meaning of the grid-size entry and any parity or seed parameter.","section":"Table I"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is potentially publishable after revision. The algebraic framework is sound and the volume-averaged validation is a strength, but the pointwise validity of the fluxes used in Sec. V needs explicit support, and the causal language in the abstract and conclusions should be tempered. The comparison issue in Fig. 2 for single energy channels should also be reworked. I do not see grounds for rejection, provided these points are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I read this as a methods paper with a physical application, and the methods part is the stronger half. The anisotropic coarse-graining implementation for Boussinesq flows is a real extension of prior work: 1D filters along k⊥ and k∥, with validation of the volume-averaged sub-scale terms against Fourier fluxes on Run I. That validation is honest and passes well, including the tricky k⊥ inverse-transfer regime near the forcing scale. The paper also correctly separates the sub-scale nonlinear terms from the buoyancy conversion term, which is a common point of confusion; their decomposition into Φu and Φθ is clean and useful.\n\nThe new physical content is the draft-conditioned local analysis: binning pointwise fluxes by |w/σw| and showing that strong draft regions have enhanced forward kinetic transfer, bidirectional potential transfer around k⊥ ≈ 15, and a parallel-direction split near kB. This is a legitimate extension of Marino et al. 2022, which only looked at global spectra and dissipation. The descriptive claim—draft regions correlate with enhanced and locally bidirectional transfer—is supported by the conditional plots. I buy that.\n\nSoft spots, in proportion: first, the causal wording. The abstract says drafts \"trigger\" transfers and the conclusions say they \"act as a mechanism.\" The evidence is conditional statistics over a single DNS run. That is a correlation with a plausible physical mechanism, not a demonstrated causal trigger. The paper needs either hedging or a mechanistic test (e.g., removing drafts, or a causality-oriented analysis).\n\nSecond, the pointwise flux validation gap is real but narrower than the stress-test note suggests. Only the volume average is validated against Fourier fluxes, and on Run I, not Run II used for the conditional statistics. The Butterworth filter can produce negative local subscale kinetic energy—the authors admit this. Whether that bias is strong enough to flip signs in the |w/σw| bins is unknown. This is a measurement-validity concern, not a fatal flaw, but it needs addressing. A few checks—e.g., comparing subdomain-averaged coarse-grained fluxes against subdomain Fourier fluxes on the same run, or testing the sensitivity to filter order—would largely close it.\n\nThird, no error bars on the central quantitative claims, and no code/data release. For a paper whose value is partly methodological, releasing the filter implementation would increase reproducibility. The run selection and time windows come from the authors' own kurtosis-based studies, which is mild circularity but not damning—the phenomenon is defined by that prior work.\n\nWho is this for? People working on stratified turbulence, LES subgrid modeling, and local energy transfer diagnostics. The method will probably get used. I would send this to review; the referee should push on the causal language and ask for pointwise validation. It deserves a serious referee despite my overall skepticism being moderate.","headline":"A genuinely useful coarse-graining toolkit for stratified flows, with draft-conditioned local flux statistics that are suggestive but not yet causal, and a validation gap at the pointwise level.","tokens_in":20560,"tokens_out":712,"would_cite":true,"duration_ms":9096,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["47.27.-i","47.27.E-"],"model":"deepseek-v4-flash","headline":"Vertical drafts in stratified turbulence act as local switches that redirect energy toward both large and small scales, strengthening kinetic-potential coupling near the buoyancy scale.","keywords":["stratified turbulence","coarse graining","energy cascade","vertical velocity drafts","intermittency","Boussinesq equations","buoyancy flux","sub-filter scale flux"],"falsifier":"Recompute the draft-conditioned flux profiles using a sharply filtered spectral flux or a different filter kernel (e.g., Gaussian) at the same cutoff wavenumbers: if the sign reversal near $k_B$ along $k_\\parallel$ and the bidirectional $\\langle S_\\theta\\rangle$ around $k_\\perp\\approx 15$ disappear, or if the enhancement no longer tracks $|w/\\sigma_w|$, the central claim is refuted. Additionally, if the fraction of points with $\\mathrm{tr}[T^{uu}]<0$ is substantial inside the high-$|w|$ bins, those local fluxes are unreliable and the claim lacks support.","tokens_in":19595,"feed_emoji":"🌊","tokens_out":6887,"duration_ms":62402,"temperature":0.7,"pith_summary":"The paper claims that in stably stratified turbulence, rare but powerful vertical velocity drafts act as local switches for the energy cascade: inside the small volume they occupy, they strengthen forward (downscale) transfer of kinetic and total energy at intermediate scales and make potential energy flow in both directions at once. This is established by refining the coarse-graining (space-filtering) method to produce pointwise estimates of cross-scale energy fluxes in simulations of the Boussinesq equations, rather than the volume-integrated Fourier fluxes used before. The authors show that in regions with $|w/\\sigma_w|$ above four, the local kinetic and total energy transfer is up to ten times the domain average, peaking near $k_\\perp\\approx 15\\approx 2k_B$, while along the parallel direction the total flux is upscale below the buoyancy wavenumber $k_B\\approx 7$ and downscale above it. These results matter because they connect large-scale intermittency, namely drafts observed in the atmosphere and oceans, to the local machinery of mixing and dissipation, suggesting where and how energy is routed across scales.","feed_headline":"Vertical drafts steer energy both up and down the cascade","feed_subtitle":"Powerful vertical drafts strengthen kinetic-potential coupling and reverse energy flow near the buoyancy scale.","key_machinery":"The machinery is the coarse-grained (space-filtered) energy budget of the Boussinesq equations, restricted to anisotropic reduced fluxes by filtering along either the perpendicular wavenumber $k_\\perp$ or the parallel wavenumber $k_\\parallel$ with a Butterworth kernel $G^{(4)}(k)=1/[1+(k/k_*)^8]$. The key identities are the sub-filter terms $S_u=-T^{uu}:\\nabla\\tilde{u}$ and $S_\\theta=-T^{\\theta u}\\cdot\\nabla\\tilde{\\theta}$, which measure kinetic and potential energy transfer through the cutoff scale $\\ell_*=1/k_*$, together with the filtered buoyancy flux $N\\tilde{\\theta}\\tilde{w}$ that converts kinetic to potential energy. Because these terms are defined pointwise in physical space, they can be averaged inside subdomains selected by $|w/\\sigma_w|$, which is what makes the local, draft-conditioned energy transfer accessible.","core_discovery":"On the paper's own terms, the central discovery is that extreme vertical velocity drafts are not passive markers of localized turbulence: they actively mediate the scale-by-scale energy budget. Conditional averages over bins of standardized vertical velocity $|w/\\sigma_w|$ show that the sub-filter kinetic flux $\\langle S_u\\rangle$ and total flux $\\langle S_{\\mathrm{tot}}\\rangle$ are increasingly forward (downscale) with draft intensity, peaking at $k_\\perp\\approx 15$, roughly twice the buoyancy wavenumber $k_B=N/U\\approx 7$, and extending from the forcing scale up to the Ozmidov scale $k_{\\mathrm{Oz}}\\approx 42$. The potential-energy flux $\\langle S_\\theta\\rangle$ becomes bidirectional in the same region, transferring potential energy simultaneously toward small and large scales, with the inversion at $k_\\perp\\approx 15$ coinciding with the maximum kinetic-to-potential conversion $N\\langle \\tilde{\\theta}\\tilde{w}\\rangle$. Along the parallel direction, the total energy flux is negative (upscale) for $k_\\parallel<k_B$ and positive (downscale) for $k_\\parallel>k_B$ in draft regions, with the upscale branch about twice as intense as the forward branch. The magnitude of all these effects grows monotonically with $|w/\\sigma_w|$ and essentially vanishes in the Gaussian core ($|w/\\sigma_w|<2.5$), so the authors conclude that vertical drafts locally trigger upscale and downscale transfers and strengthen the coupling between kinetic and potential energies at scales set by the buoyancy length.","pith_inferences":["An implication the authors leave implicit: if vertical drafts are local cascade switches, subgrid models used in weather and climate simulations might condition their eddy transfer coefficients on local vertical velocity or intermittency measures rather than on volume-averaged statistics.","Testable extension: the same conditional coarse-graining could be applied to observational velocities (ocean moorings, radar draft measurements) to look for the predicted sign reversal of the parallel flux at the buoyancy scale.","Because the effect vanishes in the Gaussian core and concentrates in about 0.6% of the volume, pointwise diagnostics of dissipation or mixing that ignore vertical-velocity conditioning will systematically underestimate the local transfer rate in draft regions.","The benchmark against Fourier fluxes was performed at $\\mathrm{Re}\\approx 97$ (Run I); applying the same volume-average comparison to a strongly intermittent run at higher Reynolds number would indicate whether the filter artifacts remain benign when the drafts are extreme."],"forward_implications":["Within draft regions ($|w/\\sigma_w|\\ge 4$), kinetic and total energy are transferred forward at up to ten times the volume-average rate, so local cascade activity is far stronger than the global mean suggests.","Along $k_\\parallel$, the total energy flux is upscale below $k_B\\approx 7$ and downscale above it, meaning vertical drafts act as a bidirectional energy mediator around the buoyancy scale.","Along $k_\\perp$, potential energy is transferred simultaneously toward large and small scales, with the sign change at $k_\\perp\\approx 15\\approx 2k_B$ coinciding with peak kinetic-to-potential conversion.","The draft-conditioned enhancement grows monotonically with $|w/\\sigma_w|$ and disappears for $|w/\\sigma_w|<2.5$, showing the effect is tied to the extreme, non-Gaussian tail of the vertical velocity.","The refined anisotropic coarse-graining reproduces the volume-averaged isotropic, parallel, and perpendicular Fourier fluxes, validating it as a local proxy for the global spectral energy transfer."],"supporting_citations":[{"why":"Establishes coarse-graining as the local-in-space measure of the energy cascade and provides the smooth-filter formalism used here.","marker":"[34]"},{"why":"Provides the space-filter derivation of the filtered energy equations that this paper adapts to Boussinesq flows.","marker":"[50]"},{"why":"Supplies the axisymmetric Fourier flux definitions and the known anisotropic flux behavior used as the volume-average benchmark.","marker":"[4]"},{"why":"Introduced the specific simulation (Run II) and the physical picture of drafts modulating local dissipation that this paper extends to scale-by-scale transfer.","marker":"[29]"},{"why":"Identified the Froude-number range and kurtosis transition for large-scale intermittency that selects the parameter regime studied.","marker":"[17]"},{"why":"Documents the multi-run parameter exploration connecting velocity and temperature bursts to small-scale intermittency, motivating the draft-conditioned analysis.","marker":"[26]"},{"why":"Explains why coarse-grained and sharp Fourier fluxes differ near the forcing scale, which the authors use to justify the benchmark agreement.","marker":"[61]"},{"why":"Earlier coarse-graining application to rotating stratified Boussinesq flows, supporting the filtered flux definitions for this system.","marker":"[39]"}],"fun_headline_variants":["Vertical drafts drive two-way energy cascade in turbulence","Extreme drafts flip energy flow at buoyancy scale","Drafts couple kinetic and potential energy locally","Strong drafts actively channel energy up and down"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the pointwise sub-filter fluxes, averaged inside vertical-velocity bins, faithfully measure local cross-scale energy transfer; the paper validates this only for volume averages, on a weaker-turbulence run, and its Butterworth filter can produce negative pointwise sub-scale kinetic energy.","fun_headline_variants_meta":{"raw":{"variants":["Vertical drafts drive two-way energy cascade in turbulence","Extreme drafts flip energy flow at buoyancy scale","Drafts couple kinetic and potential energy locally","Strong drafts actively channel energy up and down"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000551,"raw_usage":{"total_tokens":2673,"prompt_tokens":1037,"completion_tokens":1636,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":653,"completion_tokens_details":{"reasoning_tokens":1578}},"tokens_in":653,"tokens_out":1636,"duration_ms":12807,"temperature":1.0,"reasoning_tokens":1578,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T22:27:55.347231+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the draft-conditioned flux profiles using a sharply filtered spectral flux or a different filter kernel (e.g., Gaussian) at the same cutoff wavenumbers: if the sign reversal near $k_B$ along $k_\\parallel$ and the bidirectional $\\langle S_\\theta\\rangle$ around $k_\\perp\\approx 15$ disappear, or if the enhancement no longer tracks $|w/\\sigma_w|$, the central claim is refuted. Additionally, if the fraction of points with $\\mathrm{tr}[T^{uu}]<0$ is substantial inside the high-$|w|$ bins, those local fluxes are unreliable and the claim lacks support.","supporting_citations":[{"cited_title":"Pouquet, D","cited_arxiv_id":null,"evidence_quote":"Establishes coarse-graining as the local-in-space measure of the energy cascade and provides the smooth-filter formalism used here."},{"cited_title":"Meneveau and J","cited_arxiv_id":null,"evidence_quote":"Provides the space-filter derivation of the filtered energy equations that this paper adapts to Boussinesq flows."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the axisymmetric Fourier flux definitions and the known anisotropic flux behavior used as the volume-average benchmark."},{"cited_title":"Connecting Large-Scale Velocity and Temperature Bursts with Small-Scale Intermittency in Stratified Turbulence","cited_arxiv_id":"2106.07574","evidence_quote":"Introduced the specific simulation (Run II) and the physical picture of drafts modulating local dissipation that this paper extends to scale-by-scale transfer."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Identified the Froude-number range and kurtosis transition for large-scale intermittency that selects the parameter regime studied."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the multi-run parameter exploration connecting velocity and temperature bursts to small-scale intermittency, motivating the draft-conditioned analysis."},{"cited_title":"Billant and J.-M","cited_arxiv_id":null,"evidence_quote":"Explains why coarse-grained and sharp Fourier fluxes differ near the forcing scale, which the authors use to justify the benchmark agreement."},{"cited_title":"Coherent structures and spectral energy transfer in turbulent plasma: a space-filter approach","cited_arxiv_id":"1711.00291","evidence_quote":"Earlier coarse-graining application to rotating stratified Boussinesq flows, supporting the filtered flux definitions for this system."}],"review_version":1}