{"id":"4b47ac4f-f4e1-43cf-81d9-474066d940cc","arxiv_id":"2608.04325","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Mesh capture efficiency is highest when the droplet response time matches the spectrally defined flow timescale, formalized by Pi near order one.","lead":"By simulating fog droplets moving through five differently shaped wire meshes, this study proposes that capture is strongest when the droplet response time matches the time scale of the flow's unsteady fluctuations. The result is framed as design guidance, not a tested optimum, and is summarized by a new spectral matching parameter, Pi.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reported Π values and Eq. (15) parameters are mutually inconsistent; the correlation predicts near-zero capture for the droplet sizes claimed to have peak efficiency.","rationale":"The reader's weakest assumption concerned the PSD-weighted centroid and the 100–5000 Hz window; that is a legitimate input-side concern. My stress-test identified a stronger, output-side inconsistency that does not depend on speculation about the true flow timescale: substituting the published Π values (Table 5) and fit parameters (Table 6) into the published correlation (Eq. 15) yields capture efficiencies below 0.1% for the very droplets identified as having the highest capture (30 µm), while the text and η_max report about 34%. At 40 µm the prediction is essentially zero. Thus the fitted coefficients cannot have been obtained from the efficiency data in Fig. 6(b) if the Π values of Table 5 were used. This is a direct internal contradiction, not a disagreement with external consensus. The likely sources include a missing factor of 2π in τ_f, an error in α/β (e.g., values intended to be about 10× larger), a typo in Eq. (15) such as a Gaussian in Π rather than ln Π, or use of a different velocity component in Eq. (25) than reported. The reader's concern about probe placement and frequency window is related but secondary: it would explain why the Π values are what they are, but it does not address why the reported correlation and Π table cannot both be correct. Because the paper's headline conclusion—peak capture at Π≈1—is quantified by this correlation and these Π tables, the manuscript as written does not support its own central claim. A simple recomputation will confirm the inconsistency, so the paper should be revised to correct or reconcile these numbers before the spectral-dynamic framework can be adequately evaluated.","tokens_in":24564,"tokens_out":15896,"duration_ms":158358,"concrete_test":"Recompute Eq. (15) for each geometry using the Π values from Table 5 (U=5 m/s) and the α, β, η_max values from Table 6, then compare the predicted η at 30 µm and 40 µm with the simulated efficiencies in Fig. 6(b). If the predicted η deviates from the reported efficiencies by more than 10 percentage points (currently it predicts <1% where ~30% is claimed), the published parameter set is mutually inconsistent. As a follow-up, recompute f_d from Eq. (25) using the transverse instead of the streamwise velocity component to see whether the resulting Π values bring 30 µm near unity and restore a credible fit.","verdict_should_be":"REJECT","load_bearing_attack":"The central quantitative support for the Π~O(1) optimum is internally inconsistent. Table 5 reports Π at U=5 m/s for the droplet sizes claimed to span the response regimes: for the triangular mesh, Π=3.45 at 30 µm and 6.13 at 40 µm, while the text states 30 µm droplets have the highest capture efficiency and Table 6 lists η_max≈34.1%. Substituting the triangular fit (α=1.35, β=0.58) into Eq. (15) gives a Gaussian penalty exp[−1.35(ln3.45/0.58)²]≈2.1×10⁻³, so the predicted η at 30 µm is ≈0.07%; at 40 µm it is ≈2×10⁻⁶%. The same contradiction holds for every geometry (e.g., square 30 µm Π=3.29, β=0.52 gives a penalty ≈3×10⁻⁴). It is therefore impossible that Eq. (15) with Table 6 parameters was fitted to the capture-efficiency data of Fig. 6(b) if the Π values in Table 5 are the ones used. Either the Π values, the fitted α/β, the efficiency data, or the velocity-component choice feeding Eq. (25) are erroneous. Since the Π~O(1) claim rests on this correlation and these Π values, the central claim is not supported by the manuscript as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a spectral-dynamic framework for fog-harvesting mesh design, based on Eulerian–Lagrangian simulations of droplet-laden flow across five mesh geometries at a fixed inlet velocity of 5 m/s. The central quantity is the dynamic matching parameter Π = τ_p f_d, where τ_p is the droplet response time and f_d is a PSD-weighted characteristic frequency of the near-mesh flow. The authors claim that capture efficiency is maximized when Π is of order unity, that the triangular mesh outperforms the other geometries because it produces broadband, spatially distributed fluctuations, and they propose a semi-empirical correlation (Eq. 15) with geometry-dependent fitted parameters. The validation against Chen's single-cylinder collision-efficiency data is reasonable, and the manuscript is explicit about its limitations, including the single velocity and the absence of drainage modeling.","tokens_in":24876,"tokens_out":6362,"duration_ms":60715,"significance":"If the central claim were supported, the Π parameter would provide a physically motivated, geometry-sensitive alternative to the classical Stokes number for fog-harvesting mesh design, and the proposed correlation could serve as a design tool. The paper also offers a detailed spectral characterization of near-mesh unsteadiness, and the authors are transparent about the semi-empirical nature of their model and the need for future validation. However, the central claim is undermined by internal inconsistencies between the reported efficiency data, the tabulated Π values, and the fitted correlation parameters, as detailed in the major comments. The framework is interesting, but as written the evidence does not establish the Π∼O(1) optimum.","major_comments":[{"comment":"The fitted correlation is internally inconsistent with the reported Π values. For the triangular mesh, Table 6 gives η_max=34.1%, α=1.35, and β=0.58, while Table 5 lists Π=3.45 for the 30 μm droplets, which the text identifies as having the highest capture efficiency. Substituting these values into Eq. (15) gives a Gaussian penalty exp[−1.35(ln(3.45)/0.58)²] ≈ 2.1×10⁻³, so the predicted efficiency at 30 μm is about 0.07%, not 34.1%. The same contradiction holds for the square mesh (Π=3.29, β=0.52 gives a penalty of roughly 3×10⁻⁴). Thus Eq. (15) with the Table 6 parameters cannot have been fitted to the capture-efficiency data if the Table 5 Π values are the ones used. Either the Π values, the fitted α/β, the efficiency data, or the velocity-component choice feeding Eq. (25) is erroneous; as written, the correlation does not support the claim that the highest efficiency occurs at Π∼O(1).","section":"Section 3.5, Eq. (15), Tables 5 and 6"},{"comment":"The central claim that capture efficiency peaks at Π∼O(1) is contradicted by the presented data. The text states that capture efficiency increases monotonically with droplet diameter over the range considered, and Fig. 6(b) shows no decline at large diameters. Yet Section 3.2 defines Regime IV (St>69.4) with a decline in efficiency, and Section 3.4E identifies 30 μm droplets as the highest-efficiency case. Since both St and Π increase monotonically with droplet diameter, the maximum efficiency in the data occurs at the largest diameter (Π≈6 for the triangular mesh), not at Π∼1. The Π∼O(1) optimum is therefore not observable in the simulations as reported; the monotonic trend in Fig. 6(b) actively contradicts the proposed peak.","section":"Section 3.2, Fig. 6(b), Section 3.4E"},{"comment":"The characteristic frequency f_d is computed as a PSD-weighted mean over the 100–5000 Hz band, but Appendix A4 itself acknowledges that domain-scale low-frequency interactions (f<100 Hz) are poorly resolved within the 0.06 s sampling window. The paper provides no sensitivity analysis with respect to the integration band or the probe placement. If the droplet-relevant unsteadiness lies below 100 Hz, the Π values in Table 5 and the entire regime classification would change. This is a load-bearing methodological choice that needs at least a sensitivity test (e.g., recomputing f_d with a lower band limit of 16.7 Hz, the actual frequency resolution) to establish that the reported Π ordering is not an artifact of the chosen window.","section":"Appendix A4, Eq. (25), Section 3.3A"},{"comment":"There is a direct contradiction in the velocity component used for the spectral analysis. Section 3.3A states, \"The analysis focuses on the transverse velocity component, which plays a dominant role in droplet deflection and interception,\" whereas Appendix A4 states, \"The streamwise velocity component was obtained at each probe position and utilized for spectral analysis.\" Since f_d and hence Π depend on this choice, the manuscript must specify which velocity component was actually used, justify that choice physically, and, if the transverse component was intended, correct the Appendix description. Without this clarification, the reproducibility of the Π values in Table 5 is in question.","section":"Appendix A4 vs Section 3.3A"}],"minor_comments":[{"comment":"The caption refers to the \"rectangular mesh,\" but the text and the geometry description use \"square mesh\"; the terminology should be unified throughout.","section":"Fig. 4(b) caption"},{"comment":"Reference 26 is cited near Eq. (10) and in the text as Langmuir's paradigm, but the reference list entry for item 26 is Johnson, Kendall, and Roberts (1971). Please check and correct the citation numbering.","section":"References"},{"comment":"There are spelling errors in key passages: \"persistant\" should be \"persistent\" (Section 3.5), and \"remaines\" should be \"remains\" (Conclusions).","section":"Section 3.5 and Conclusions"},{"comment":"The appendix figure numbering is inconsistent with the in-text references (e.g., Table 4 refers to Fig. A8 for the square mesh, but the square mesh appears in Appendix A7). Please align the figure numbering.","section":"Appendix A5–A8"},{"comment":"The Stokes-number thresholds for Regimes III and IV (44.4 and 69.4) are presented without derivation or a table showing the corresponding droplet sizes; please indicate how these thresholds are obtained and how they relate to Fig. 6(b).","section":"Section 3.2"}],"recommendation":"reject","confidential_remarks":"The manuscript contains a substantial simulation effort and an interesting conceptual framework, but the internal inconsistencies—particularly the contradiction between Eq. (15) with Table 6 and the Π values in Table 5, and the monotonic efficiency trend in Fig. 6(b) versus the claimed Π∼O(1) peak—are load-bearing and cannot be resolved by minor revision. The authors would need to either provide new simulations or substantially reframe the central claim. I therefore recommend rejection, though the topic is within the journal's scope and the authors' emphasis on spectral, geometry-dependent flow timescales is worth pursuing in a future, carefully validated study."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's the read on arXiv:2608.04325. The paper does something genuinely new: it moves fog-mesh analysis from mean-flow and Stokes-number thinking to a frequency-domain characterization, defining Pi = tau_p f_d from a PSD-weighted spectral centroid. The simulation campaign is respectable, with a proper single-cylinder validation, grid-independence study, and an honest set of caveats. That part deserves credit.\n\nThe trouble is that the central quantitative claim, the Pi~O(1) optimum and its fitted correlation Eq. (15), does not survive contact with the paper's own tables. Table 5 gives Pi at U=5 m/s: for the triangular mesh, 30 um droplets have Pi=3.45 and 40 um have Pi=6.13. Table 6 gives eta_max=34.1%, alpha=1.35, beta=0.58. Plugging those into Eq. (15) gives eta at 30 um of about 0.07% and at 40 um about 2e-6%. But Fig. 6(b) shows capture efficiency increasing monotonically with droplet diameter, and the text explicitly says 30 um droplets give the highest capture efficiency. So either the Pi values, the fitted alpha/beta, or the efficiency data are wrong. The same contradiction repeats for every geometry. This isn't a footnote-level slip; it is the quantitative basis for the paper's main selling point. As written, the optimum is not supported.\n\nThere are also smaller inconsistencies. Section 3.3 says the spectral analysis uses the transverse velocity component, while Appendix A4 says streamwise. The 0.06 s sampling window with a 100 Hz lower bound is another soft spot: the paper itself admits low-frequency content is poorly resolved, but that's exactly where wake and domain-scale structures would live. And the correlation is fitted to the same simulated data that defines Pi, which the authors honestly call semi-empirical; that limits the predictive claim.\n\nThe reason I'd still send this to a serious referee is that the core idea, that geometry shapes the flow timescale spectrum and droplet capture responds to that timescale, is testable and potentially useful for design. The simulation work appears honestly done. But the manuscript in its current form is internally inconsistent at its load-bearing point. The authors need to recheck Table 5, Table 6, and Fig. 6 together, then refit or redraw. If the data are as presented, the monotonic trend contradicts the Gaussian penalty, and the Pi~O(1) claim needs a different representation or additional high-Pi simulations showing the downturn.\n\nWho benefits: researchers working on fog harvesting or inertial particle capture in porous structures. It's a desk-review boundary case: valuable idea, flawed execution. If the internal inconsistency is fixed, this could make a solid contribution.","headline":"Promising spectral-dynamic framing, but the paper's central Pi~O(1) optimum is contradicted by its own Table 5, Table 6, and Fig. 6(b).","tokens_in":25396,"tokens_out":3043,"would_cite":false,"duration_ms":29493,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["47.55.Kf"],"model":"deepseek-v4-flash","headline":"Fog-harvesting meshes capture the most droplets when the droplet response time matches the mesh-induced flow timescale, a matching condition that explains why a triangular mesh beats square, trapezoidal, hexagonal, and harp-shaped designs.","keywords":["fog harvesting","droplet capture efficiency","spectral analysis","Eulerian-Lagrangian simulation","dynamic timescale matching","mesh geometry optimization","power spectral density","fog droplet size distribution"],"falsifier":"A decisive check is to recompute $f_d$ from probe records long enough to resolve frequencies down to about 1 Hz at the same simulation setup and test whether the geometry ranking of $\\Pi$ still matches the capture-efficiency ranking; if efficiencies peak at $\\Pi$ values far from unity under the wider window, or if the ordering changes, the spectral-centroid definition of $\\tau_f$ is the refuted link. A complementary wind-tunnel test with monodisperse droplets and triangular versus square meshes at several wind speeds would check the stronger prediction that the efficiency peak tracks the $\\Pi \\approx 1$ contour as $f_d$ scales with velocity.","tokens_in":24340,"feed_emoji":"💧","tokens_out":21722,"duration_ms":181650,"temperature":0.7,"pith_summary":"Fog-harvesting efficiency is usually discussed through mean-flow metrics, but this paper claims that what really decides whether a droplet hits a mesh fiber is whether the droplet can keep time with the flow's own fluctuations. Using two-way coupled Eulerian–Lagrangian simulations of 2–40 µm fog droplets across five mesh geometries, the authors extract a characteristic flow timescale from the power spectrum of near-mesh velocity and compare it with the droplet response time. The resulting matching parameter, $\\Pi = \\tau_p f_d$, gives the highest capture efficiency when it is of order unity; geometries with broadband, spatially balanced unsteadiness (triangular) sustain that condition, while narrowband, weak, or intermittent spectra (square, trapezoidal, hexagonal, vertical) degrade it. This yields a mechanistic explanation for the observed geometry ranking and a semi-empirical correlation, recasting mesh design as a timescale-matching problem rather than a search for stronger fluctuations. Because drainage and re-entrainment are not modeled, the reported numbers are capture efficiencies, not field collection efficiencies.","feed_headline":"Fog meshes catch most when droplets and airflow fall into sync","feed_subtitle":"Simulations show capture efficiency peaks at one matching condition, a concrete design target for mesh shape.","key_machinery":"The load-bearing object is the dynamic matching parameter $\\Pi = \\tau_p f_d$, which combines the Stokes-regime droplet response time $\\tau_p = \\rho_d D_d^2/(18\\mu)$ with the characteristic flow frequency $f_d$ extracted from the power spectral density of near-mesh velocity fluctuations. $f_d$ is computed as the PSD-weighted mean frequency (the spectral centroid) over the 100–5000 Hz band from 0.06 s probe signals, a window sized to capture fiber-scale vortex shedding near 5000 Hz. The parameter does the work that the classical Stokes number cannot: because inlet velocity, strand width, and shade coefficient are fixed across the five geometries, $St$ is identical for a given droplet size, whereas $\\Pi$ inherits each geometry's spectral signature and orders the configurations. The companion correlation $\\eta = \\eta_{\\max}\\frac{\\Pi}{\\Pi+\\Pi^{-1}}\\exp[-\\alpha(\\ln\\Pi/\\beta)^2]$ encodes the three interaction regimes — flow-following ($\\Pi \\ll 1$), dynamically matched ($\\Pi \\sim O(1)$), and inertia-dominated ($\\Pi \\gg 1$) — with a Gaussian penalty for timescale mismatch in logarithmic space.","core_discovery":"The central claim is that droplet capture in fog-harvesting meshes is controlled by dynamic compatibility between droplet inertia and geometry-induced flow unsteadiness, not by fluctuation magnitude alone. The paper defines a characteristic flow frequency $f_d$ as the PSD-weighted mean frequency (spectral centroid) of near-mesh velocity fluctuations, giving a flow timescale $\\tau_f \\sim 1/f_d$, and compares it with the Stokes-regime droplet response time $\\tau_p = \\rho_d D_d^2/(18\\mu)$ through the parameter $\\Pi = \\tau_p/\\tau_f = \\tau_p f_d$. $\\Pi$ is presented as a frequency-based analogue of the Stokes number: at fixed inlet velocity, strand width, and shade coefficient, the classical $St$ cannot distinguish the geometries for a given droplet size, whereas $\\Pi$ inherits each geometry's spectral signature and orders the five configurations. Capture efficiency is maximized when $\\Pi$ is of order unity (roughly 1–6), and the triangular mesh wins because its broadband, moderately amplified spectrum extends the region over which droplets experience sustained, matched forcing. A physics-inspired correlation, $\\eta = \\eta_{\\max}\\frac{\\Pi}{\\Pi+\\Pi^{-1}}\\exp[-\\alpha(\\ln\\Pi/\\beta)^2]$, with geometry-dependent fitted parameters (reported $R^2$ from 0.936 to 0.981), collapses the diameter-dependent efficiency curves, and the authors frame the result as mechanistic design guidance rather than a proven universal optimum, noting the single-velocity, uniform-inflow conditions simulated.","pith_inferences":["The matching criterion implies a screening protocol the paper does not run: at a target wind speed, measure or simulate the near-mesh velocity spectrum of candidate geometries, compute each spectral centroid $f_d$, and pick the geometry that brings the mass-dominant droplet size to $\\Pi \\approx 1$.","If the expected scaling $f_d \\propto U/l_f$ holds, $\\Pi$ acts as a reduced frequency and mesh design becomes a resonance-matching problem: a mesh tuned for one wind speed could be retuned for another by rescaling strand width, a quantitative prediction that multi-velocity experiments could test.","The paper's own analysis window leaves an acknowledged blind spot below 100 Hz: the 0.06 s sampling resolves fiber-scale shedding but not domain-scale or atmospheric low-frequency unsteadiness, so the framework's field relevance depends on those low frequencies being unimportant for droplet interception.","A reader should note an internal inconsistency in the manuscript: Section 3.3 says the spectral analysis focuses on the transverse velocity component, while Appendix A4 says the streamwise component was analyzed; because $f_d$ enters every $\\Pi$ value and the final correlation, resolving which component defines the timescale would settle what the matching parameter measures."],"forward_implications":["At the simulated conditions (5 m/s inlet, 0.6 shade coefficient), capture efficiency ranks triangular above square above trapezoidal above hexagonal above vertical for all droplet sizes, and the ranking follows from how each geometry shapes the spectral distribution of near-mesh unsteadiness.","The framework reduces mesh design to a single objective: maximize the spatial and spectral extent over which $\\Pi \\approx 1$ holds for the droplet sizes that carry most liquid water mass, rather than maximizing fluctuation intensity.","$\\Pi$ is flow-dependent, not geometry-static: if Strouhal-number similarity holds ($f_d \\propto U/l_f$), the same mesh can be well-matched at one wind speed and poorly matched at another, so a given geometry must be chosen for its intended operating envelope.","Because the simulations assume deterministic adhesion of low-Weber-number droplets and exclude drainage, the reported values are capture efficiencies (aerodynamic times deposition) and would be multiplied by a drainage efficiency of roughly 45–95% to approach field collection performance.","The paper offers Eq. (15) with geometry-dependent fitted parameters as a physics-inspired semi-empirical model whose predictive use requires an independent estimate of $f_d$, for instance from Strouhal-number correlations, together with multi-velocity and turbulent-inflow validation."],"supporting_citations":[{"why":"Supplies the single-cylinder collision-efficiency benchmark against which the solver's capture predictions are validated, with agreement within 1.4–3.2%.","marker":"[10]"},{"why":"Supplies the aerodynamic/deposition/drainage decomposition of collection efficiency and the wire-scale Reynolds-number basis used to set the characteristic length $l_f$.","marker":"[14]"},{"why":"Provides the trajectory-based inertial-impaction data and the K–φ parameterization that the validation case reproduces.","marker":"[28]"},{"why":"Provides the in-situ fog droplet size distributions used to calibrate the two-parameter inlet droplet spectrum, with peaks between 11 and 25 µm.","marker":"[32]"},{"why":"Supplies the Welch-averaged spectral estimation procedure that produces the power spectra from which the characteristic frequency $f_d$ is extracted.","marker":"[47]"},{"why":"Defines the Euler–Lagrange coupling strategy, including the requirement that computational cells remain at least twice the particle diameter for physical fidelity.","marker":"[9]"},{"why":"Supplies the drag-coefficient correlation used to advance droplet motion in the Lagrangian tracking.","marker":"[41]"},{"why":"Supplies the generalized unsteady force equation for a sphere that governs droplet translation in the dispersed-phase model.","marker":"[38]"}],"fun_headline_variants":["Droplet capture peaks when flow and droplet timescales match","Why fog meshes work best: a frequency match between droplets and air","Triangular meshes win by matching droplet inertia to flow unsteadiness","Capture efficiency hinges on droplet-flow spectral resonance","Fog mesh efficiency peaks when droplets sync with unsteady airflow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the spectral centroid $f_d$ computed from 0.06-second probe signals over the 100–5000 Hz band captures the flow timescale that actually controls droplet capture; if the controlling unsteadiness lies below 100 Hz, or if a single weighted mean cannot represent a broadband forcing spectrum, then the $\\Pi$ ordering and the order-unity condition are artifacts of the chosen frequency window.","fun_headline_variants_meta":{"raw":{"variants":["Droplet capture peaks when flow and droplet timescales match","Why fog meshes work best: a frequency match between droplets and air","Triangular meshes win by matching droplet inertia to flow unsteadiness","Capture efficiency hinges on droplet-flow spectral resonance","Fog mesh efficiency peaks when droplets sync with unsteady airflow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000772,"raw_usage":{"total_tokens":3506,"prompt_tokens":1123,"completion_tokens":2383,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":739,"completion_tokens_details":{"reasoning_tokens":2310}},"tokens_in":739,"tokens_out":2383,"duration_ms":15702,"temperature":1.0,"reasoning_tokens":2310,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T19:45:06.781069+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive check is to recompute $f_d$ from probe records long enough to resolve frequencies down to about 1 Hz at the same simulation setup and test whether the geometry ranking of $\\Pi$ still matches the capture-efficiency ranking; if efficiencies peak at $\\Pi$ values far from unity under the wider window, or if the ordering changes, the spectral-centroid definition of $\\tau_f$ is the refuted link. A complementary wind-tunnel test with monodisperse droplets and triangular versus square meshes at several wind speeds would check the stronger prediction that the efficiency peak tracks the $\\Pi \\approx 1$ contour as $f_d$ scales with velocity.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the single-cylinder collision-efficiency benchmark against which the solver's capture predictions are validated, with agreement within 1.4–3.2%."},{"cited_title":"Water Resour Manag 31(1)","cited_arxiv_id":null,"evidence_quote":"Supplies the aerodynamic/deposition/drainage decomposition of collection efficiency and the wire-scale Reynolds-number basis used to set the characteristic length $l_f$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the trajectory-based inertial-impaction data and the K–φ parameterization that the validation case reproduces."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the in-situ fog droplet size distributions used to calibrate the two-parameter inlet droplet spectrum, with peaks between 11 and 25 µm."},{"cited_title":"Renew Energy 119","cited_arxiv_id":null,"evidence_quote":"Supplies the Welch-averaged spectral estimation procedure that produces the power spectra from which the characteristic frequency $f_d$ is extracted."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Euler–Lagrange coupling strategy, including the requirement that computational cells remain at least twice the particle diameter for physical fidelity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the drag-coefficient correlation used to advance droplet motion in the Lagrangian tracking."},{"cited_title":"Generalized Basset –Boussinesq–Oseen equation for unsteady forces on a sphere in a compressible flow","cited_arxiv_id":null,"evidence_quote":"Supplies the generalized unsteady force equation for a sphere that governs droplet translation in the dispersed-phase model."}],"review_version":1}