{"id":"f5e9a657-eb05-4d90-8cd7-fd0e839f1edd","arxiv_id":"2611.14705","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Geometry-specific LES pressures and analytical fluctuation corrections improve airflow network predictions of natural ventilation, increasing R² from ~0.4 to ~0.9 for windows.","lead":"This paper tests two improvements to airflow network models used to predict natural ventilation in buildings: using geometry-specific exterior pressures from large eddy simulations instead of standard ASHRAE coefficients, and adding analytical corrections for pressure fluctuations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"Quasi-steady assumption not verified for the specific flows; the fluctuation corrections rely on it, but the paper provides no direct test that dynamic effects are negligible at the relevant timescales for these building openings.","rationale":"The reader's weakest_assumption correctly identifies the quasi-steady assumption as load-bearing. My analysis agrees and sharpens the concern: the paper does not test this assumption against the LES data, even though the data exist to do so. The Monte Carlo verification is not a substitute because it assumes quasi-steady behavior. The paper's own acknowledgments (sentences [b122], [b127], [b139]) show awareness of the assumption's limitations but do not resolve them. This is not a fatal flaw—the assumption is plausible and supported by prior work—but it is the single most load-bearing concern because if it fails, the entire fluctuation correction framework collapses. The concrete test I propose is feasible with the existing data and would directly settle the issue. I recommend keeping the CONDITIONAL verdict because the concern is significant but not disqualifying; the paper's other contributions (geometry-specific pressures) are robust. The agreement_with_reader is 'agree' because we identify the same weakest assumption.","tokens_in":21974,"tokens_out":679,"duration_ms":8340,"concrete_test":"For a subset of the LES cases (e.g., one wind direction, one canopy density, one ventilation configuration), compute the instantaneous ventilation rate predicted by the orifice equation using the full time-resolved pressure difference from the exterior-only LES. Compare this predicted instantaneous rate to the actual instantaneous ventilation rate from the coupled interior-exterior LES. If the RMSE of the instantaneous prediction is small relative to the mean ventilation rate, the quasi-steady assumption is supported. If the RMSE is comparable to the fluctuation correction magnitude (10–40%), the assumption is not justified and the fluctuation corrections may be unreliable.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim that fluctuation corrections reduce RMSE by 10–40% depends entirely on the quasi-steady assumption (Section 3.2, paragraph 2). The authors cite Etheridge (2000) for support, but that reference addresses idealized openings, not the specific window and skylight geometries in an urban canopy. The paper acknowledges that 'real openings need not satisfy either premise exactly' (sentence [b139]) and that the quasi-steady assumption 'requires the full time series' (sentence [b127]), but it never directly tests whether inertial or compressibility effects are small at the fluctuation timescales present in the LES data. Without this check, the fluctuation corrections could be compensating for dynamic effects rather than correctly modeling quasi-steady nonlinearity. The paper's own Monte Carlo verification (Section 3.3) uses synthetic Gaussian fluctuations, which by construction satisfy the quasi-steady assumption, so it does not validate the assumption for real flows. The LES data could be used to test this: compare the instantaneous orifice-equation prediction (using the full pressure time series) against the actual instantaneous ventilation rate from the coupled LES. If the quasi-steady assumption holds, the two should agree closely; if not, the discrepancy reveals the magnitude of neglected dynamic effects.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper proposes two improvements to airflow network models for natural ventilation: (i) using geometry-specific surface pressures from exterior-only LES instead of standard ASHRAE pressure coefficients, and (ii) analytical corrections for the nonlinear effects of pressure fluctuations on mean ventilation rates. The authors evaluate these improvements against a large dataset of coupled interior-exterior LESs (5824 windows, 2688 skylights, 2752 interiors) across varying wind speeds, directions, temperature differences, canopy densities, and interior layouts. They report that geometry-specific pressures reduce RMSE by approximately 50%, and fluctuation corrections further reduce RMSE by 10–40% for most opening types, increasing R² from ~0.4 to ~0.9 for windows. The fluctuation corrections are derived analytically from the orifice equation, verified with Monte Carlo simulations, and validated against the LES data. The paper also discusses practical implementation options (bulk vs. opening-specific fluctuation intensities, in-network vs. post-network corrections) and recommends pressure-based corrections with bulk values as a practical alternative.","tokens_in":22300,"tokens_out":2042,"duration_ms":13472,"significance":"If the results hold, this work would represent a significant advance in the accuracy of airflow network models for natural ventilation, which are widely used in building energy simulation. The 50% RMSE reduction from geometry-specific pressures addresses a known limitation of standard ASHRAE coefficients in urban canopies. The fluctuation corrections provide a theoretically grounded way to account for turbulence effects that are typically neglected. The large validation dataset and the systematic comparison of implementation options are strengths. The paper also provides open data and code, supporting reproducibility. The main significance is in demonstrating that exterior-only LES can provide sufficient information to substantially improve ventilation predictions, which could enable more reliable design of natural ventilation in urban settings.","major_comments":[{"comment":"The quasi-steady assumption is load-bearing for the entire fluctuation correction derivation, but the paper does not directly test its validity for the specific flows in the LES dataset. The authors cite Etheridge (2000) for support and acknowledge that 'real openings need not satisfy either premise exactly' (sentence [b139]), but they do not quantify the magnitude of dynamic effects (inertia, compressibility) at the relevant timescales for these window and skylight geometries in the urban canopy. The Monte Carlo verification (Section 3.3) uses synthetic Gaussian fluctuations that by construction satisfy the quasi-steady assumption, so it does not validate the assumption for real flows. The LES data could be used to test this: compare the instantaneous orifice-equation prediction (using the full pressure time series from exterior-only LES) against the actual instantaneous ventilation rate from the coupled LES. If the quasi-steady assumption holds, the two should agree closely; if not, the discrepancy reveals the magnitude of neglected dynamic effects. Without this check, the reported RMSE reductions from fluctuation corrections could be partly compensating for dynamic effects rather than correctly modeling quasi-steady nonlinearity. This is a major concern because the central claim that fluctuation corrections reduce RMSE by 10–40% depends entirely on the validity of the quasi-steady assumption for these flows.","section":"Section 3.2, paragraph 2; Section 4.3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The discharge coefficients C_d are calibrated to minimize RMSE against the LES data separately for each model formulation. This means the comparison between models is not a pure test of the pressure source or fluctuation correction, because different C_d values are used for different models. The authors acknowledge this for the ASHRAE baseline (using 0.611) and note that the LES mean-pressure baseline also uses 0.611, but the fluctuation-corrected models use C_d values around 0.7. The paper argues that this is appropriate because C_d should be calibrated for each model, but it conflates the effect of the fluctuation correction with the effect of changing C_d. To isolate the improvement from fluctuation corrections, one would need to use the same C_d (e.g., 0.611) for both the mean-pressure and fluctuation-corrected models and compare their performance. The current comparison may overstate the benefit of fluctuation corrections because the C_d values are optimized separately. The authors should present results with a fixed C_d (e.g., 0.611) for all models to show the net effect of the fluctuation correction alone, in addition to the calibrated results.","section":"Section 3.5.2, Table 3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper reports that fluctuation corrections reduce RMSE by 10–40% for most opening types and ventilation configurations, but the figures show that for some categories (e.g., skylight inflow, single-sided rooms with skylights), the fluctuation corrections increase RMSE or bias. The text acknowledges this briefly ('while predictions for skylight inflows generally become less accurate'), but the overall narrative emphasizes the improvements. The authors should provide a more balanced discussion of when the corrections help versus hurt, and quantify the frequency and magnitude of degradation. This is important for practitioners deciding whether to apply these corrections.","section":"Section 4.3.1, Figures 5 and 6","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The bulk fluctuation intensities (sigma_n) are measured from the LES data and then used in the fluctuation corrections. This introduces a potential circularity: the corrections are evaluated on the same data from which the sigma values are derived. The authors acknowledge this in the reader's take (circularity score 3.0) but do not discuss it in the paper. While the correction form itself is derived analytically and verified with Monte Carlo, the specific sigma values used in the validation are data-dependent. The paper should discuss how the results might change if sigma were estimated from a different source (e.g., empirical correlations or simpler models) and whether the reported RMSE reductions are robust to uncertainty in sigma.","section":"Section 3.4.1, Table 2","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The abstract states 'reduces RMSE by approximately 50%' and 'further reduce RMSE by 10–40%'. It would be helpful to specify the baseline for these percentages (e.g., relative to the ASHRAE mean-pressure model).","section":"Abstract","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The shelter classes and factors are described but the conversion from ASHRAE pressure coefficients (referenced to H_R) to U_10 is not fully explained. The logarithmic profile with roughness length 0.3 m is mentioned, but the value of H_R (room height) is not given. Please provide H_R or the conversion factor used.","section":"Section 2.2.1, Table 1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The transition point I* is given as 1/sqrt(3) for pressure and 1/sqrt(2) for ventilation. The derivation of these specific values is not explained in the main text; it appears to come from the tangency condition in Appendix A.2. Please briefly justify or cite the appendix equation.","section":"Section 3.2.4, Equation (19)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The synthetic data generation uses normal distributions for both ventilation rates and pressure differences. However, the orifice equation is nonlinear, so both cannot be normally distributed simultaneously. The authors acknowledge this in Section 3.2.2 ('Because the orifice relation is nonlinear, Δp' and q' cannot both be symmetric'), but the verification uses each separately. This is fine, but the text should clarify that the two experiments are independent and each tests one formulation.","section":"Section 3.3.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The scatter plots show R² values for inflow windows, outflow windows, and interiors. It would be helpful to also report the R² for the ASHRAE model on the same plot for direct comparison, rather than only in the text.","section":"Section 4.2, Figure 7","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The recommended C_d = 0.7 for fluctuation-corrected models is based on the specific geometries studied. The paper should more explicitly state that this value may not be transferable and that users should calibrate for their own geometries if possible.","section":"Section 5.2.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper uses the term 'mean-pressure' to refer to the standard airflow network model that uses time-averaged pressures. This is clear, but occasionally the term 'mean-pressure' could be confused with 'mean pressure difference'. Consider using 'time-averaged pressure' or 'steady pressure' for clarity.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The closed-form expressions for H under Gaussian fluctuations are useful. However, the paper does not discuss how non-Gaussian fluctuations (which are common in turbulent flows) would affect the corrections. A brief comment on the sensitivity to the fluctuation distribution would strengthen the analysis.","section":"Appendix A.3","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"Using flow-resolved building pressures and closed-form fluctuation corrections, airflow-network models predict natural ventilation rates near LES accuracy from exterior-only pressure data.","keywords":["natural ventilation","airflow networks","large eddy simulation","urban canopy flow","pressure coefficients","discharge coefficient","fluctuation correction","building energy modeling"],"falsifier":"Run a sharp-edged opening in a laboratory model or full-scale building under a turbulent incident flow at fixed fluctuation intensity, then sweep the fluctuation frequency upward: the quasi-steady correction predicts the mean rate depends only on the pressure distribution and intensity I, not on frequency, so any systematic shift of the mean rate with frequency (an inertia or compressibility signature) would falsify the correction outside the quasi-steady regime.","tokens_in":25398,"feed_emoji":"🌬️","tokens_out":1890,"duration_ms":165693,"temperature":0.7,"pith_summary":"Building energy models predict natural ventilation by solving airflow networks of orifice equations fed with time-averaged exterior pressures, and those predictions carry two persistent errors: the pressures come from isolated-building wind-tunnel coefficients, and the mean-pressure treatment ignores that ventilation scales with the square root of a fluctuating pressure difference. This paper claims both errors can be fixed, and that fixing them suffices to bring predicted mean ventilation rates close to measurement quality. Replacing standard ASHRAE coefficients with geometry-specific surface pressures extracted from exterior-only large-eddy simulations cuts the RMSE of predicted ventilation rates by roughly 50%, because the simulations resolve urban canopy interference that shelter-factor adjustments miss. The paper then derives analytical corrections for the square-root (fluctuation) error in two regimes, mean-dominated and fluctuation-dominated, and joins them into a parameter-free piecewise model that it validates against coupled interior-exterior LESs of thousands of windows, skylights, and interior layouts. Together the two improvements raise the coefficient of determination for window ventilation from about 0.4 to about 0.9.","feed_headline":"Geometry-specific pressures cut ventilation error in half","feed_subtitle":"Analytical fluctuation corrections lift R2 from 0.4 to 0.9 for windows.","key_machinery":"The central object is the orifice equation q = S C_d A $\\sqrt$(2|$\\Delta$ p|/rho), applied instantaneously (quasi-steady) rather than in its mean-pressure form, and the piecewise fluctuation-correction factor F_fc(I) that multiplies the mean-pressure rate: F_fc = $\\sqrt$(1-$I^{2}$) for I < I^* and F_fc = 1/(2H_T) for I >= I^*, where I is the fluctuation intensity (the ratio of fluctuation standard deviation to the mean rate or pressure) and H_T is a tangent-fitted low-order mean of fluctuation magnitudes. The transition points I^* = 1/$\\sqrt$(3) (pressure form) and I^* = 1/$\\sqrt$(2) (ventilation form) are set by requiring tangential intersection of the two branches, so the model carries no fitted transition parameter. Two complementary machineries work together: exterior-only LES supplies geometry-specific mean surface pressures, and also the pressure-form fluctuation intensities sigma_{$\\Delta$ C_p}, while the piecewise correction removes the systematic nonlinearity that mean-pressure models leave behind, with a recommended discharge coefficient of about C_d = 0.7 for sharp-edged square openings.","core_discovery":"The central claim is that the two long-standing inaccuracies of airflow-network ventilation models are independent and each has a tractable fix. First, aerodynamic pressure coefficients evaluated at the actual openings from exterior-only LES, replacing the ASHRAE isolated-building coefficients with shelter factors, reduce RMSE of predicted ventilation rates by about 50%, with the gain coming mostly from capturing urban canopy flow; the exterior-only and coupled interior-exterior simulations produce nearly identical pressure fields away from the openings, so the cheaper exterior-only run suffices. Second, the residual systematic nonlinearity between mean-pressure predictions and measured rates is the square-root error: because the ventilation rate scales as the square root of the pressure difference, pressure fluctuations bias the time-averaged rate. Under a quasi-steady assumption the correct mean rate is q approx q_MP $\\sqrt$(1-$I^{2}$) when fluctuations are small (mean-dominated) and q approx q_MP/(2H) when they dominate, where I is a fluctuation intensity and H a low-order mean of fluctuation magnitudes; a smooth piecewise model spanning both regimes, with the transition fixed by tangency rather than by a fitted parameter, removes the systematic nonlinearity and cuts RMSE a further 10-40% for most opening types. The combined model, validated against coupled LESs of 5824 windows, 2688 skylights, and 2752 interior sections across idealized urban canopies, lifts $R^{2}$ for windows from about 0.4 to about 0.9, showing that accurate mean ventilation predictions need only exterior-flow pressure information plus the analytical correction.","pith_inferences":["Since geometry-specific pressures account for the largest share of the improvement (~50% RMSE) while fluctuation corrections add 10-40%, further accuracy gains will likely come from cheaper or better exterior pressure prediction, such as surrogates or data-driven pressure fields, rather than from refining fluctuation models.","The observed near-equality of exterior-only and coupled pressure fields away from the openings suggests a general decoupling principle: exterior-only CFD or field measurements can feed airflow networks without ever solving the interior, a workflow extension beyond the specific geometries tested here.","Because the fluctuation-dominated correction assumes symmetric fluctuations, strongly skewed real urban pressure signals would violate that premise; the paper's own tangent-fixed transition would then need recalibration, a testable extension.","Historically reported reductions in discharge coefficient under cross-flow may partly be fluctuation artifacts rather than pure physical changes in C_d, since uncorrected calibrations absorb the square-root error into the fitted coefficient; the paper hints at this and formalizing the split is a further interpretive step."],"forward_implications":["Building energy models can reach roughly R^2 = 0.9 for window ventilation using only exterior-flow pressure information, so a coupled interior-exterior simulation is not needed for accurate mean rates.","A discharge coefficient near 0.7, rather than the potential-flow value 0.611, is recommended for fluctuation-corrected sharp-edged square openings, and coefficients calibrated without fluctuation corrections should not be reused inside corrected models.","Bulk fluctuation intensities and post-network implementation perform nearly as well as opening-specific, in-network corrections, so existing airflow-network solvers can adopt the correction without rewriting their internal mass-balance solve.","Pressure-based corrections are preferred over ventilation-based ones because the required fluctuation intensity comes from exterior-only data, consistent with the recommended exterior-only workflow.","In the fluctuation-dominated regime the effective pressure scaling shifts from q ~ sqrt(Delta p) to q ~ Delta p, a change that an uncorrected mean-pressure calibration could misread as a viscous effect."],"supporting_citations":[{"why":"Supplies the coupled interior-exterior LES dataset that the airflow network predictions are validated against, including canopy, wind, and temperature variations.","marker":"Bachand et al. (2026)"},{"why":"Provides the baseline isolated-building pressure coefficients and shelter classes that geometry-specific pressures are compared against and replace.","marker":"ASHRAE (2025)"},{"why":"Supports the quasi-steady assumption for natural ventilation openings, which the entire fluctuation-correction derivation rests on.","marker":"Etheridge (2000)"},{"why":"The source of the mean-dominated fluctuation correction, which the paper shows reduces to this earlier pulsating-flow-meter result.","marker":"Mottram (1971)"},{"why":"Supplies the orifice equation, the potential-flow derivation of C_d = 0.611, and the airflow-network framework the paper builds on.","marker":"David Etheridge (2012)"},{"why":"Defines the shelter classes and factors used to adjust ASHRAE coefficients in the baseline model.","marker":"Walker and Wilson (1998)"},{"why":"The wind-tunnel work underlying the ASHRAE pressure coefficients used as the baseline boundary condition.","marker":"Holmes (1994)"}],"fun_headline_variants":[],"cache_read_input_tokens":0,"weakest_assumption_plain":"The entire fluctuation-correction chain rests on the quasi-steady assumption that inertia and compressibility are negligible at the timescale of pressure fluctuations, so the square-root orifice relation holds instantaneously; if real openings show dynamic effects at those timescales, the closed-form corrections would not apply.","fun_headline_variants_meta":{"raw":{"variants":[]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.00011,"raw_usage":{"cost":0.000110201,"is_byok":false,"cost_details":{"upstream_inference_cost":0.000110201,"upstream_inference_prompt_cost":0.000109837,"upstream_inference_completions_cost":3.64e-07},"total_tokens":1209,"prompt_tokens":1207,"completion_tokens":2,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":0,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":0}},"tokens_in":1207,"tokens_out":2,"cost_source":"openrouter_usage","duration_ms":1776,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":0,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-11-13T03:21:28.857991+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":"Run a sharp-edged opening in a laboratory model or full-scale building under a turbulent incident flow at fixed fluctuation intensity, then sweep the fluctuation frequency upward: the quasi-steady correction predicts the mean rate depends only on the pressure distribution and intensity I, not on frequency, so any systematic shift of the mean rate with frequency (an inertia or compressibility signature) would falsify the correction outside the quasi-steady regime.","supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}