{"id":"fbc0c6e8-0c30-40be-a930-447ed7d35c57","arxiv_id":"2411.16223","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"A coupled PDE model of crowd flow, SEISV infection spread, and potential-flow ventilation is used to numerically compare macro control measures for epidemics transport in indoor evacuations.","lead":"This paper builds a computer model that couples moving crowds, disease spread, and ventilation airflow to test how control measures change the number of people exposed during an evacuation. It finds that stronger ventilation, higher walking speed, more spacing, and masks or vaccines all reduce predicted exposure, but the best pattern depends on the room and exit layout.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (19), the infection-coefficient transport equation, is not a well-posed IBVP: no boundary conditions are specified and the source term has inconsistent units, so the ventilation and exposure comparisons are not fully determined.","rationale":"I agree with the reader that the potential-flow model for ventilation is a major simplification and that the paper would be strengthened by comparing against a more realistic indoor airflow model. However, the single most load-bearing concern is more basic: the infection-coefficient equation itself is not fully specified. Equations (15)-(18) and (19) form the epidemiological core that converts local infected density and ventilation into exposure, and every reported trend depends on beta(x,t). Without boundary conditions for Eq. (19), the numerical tests in Section 4 cannot be uniquely reproduced, and without dimensional consistency, the quantitative values of exposed percentages are not physically interpretable. This is not an external calibration issue; it is an internal well-posedness issue. The paper deserves credit for a coherent coupled framework, a stable numerical scheme, and several non-obvious findings such as the reversal of ventilation-direction benefits in large crowds. Those findings may survive after the equation is closed correctly, but the current manuscript does not provide enough information to verify that they do. A conditional verdict is appropriate: the authors should supply the missing boundary conditions, fix the units, and demonstrate that the qualitative trends are robust under reasonable alternative closures. The reader's weakest-assumption pick captures a real limitation, but the boundary/unit problem in Eq. (19) is more direct and more consequential for the central claim.","tokens_in":18259,"tokens_out":5642,"duration_ms":194092,"concrete_test":"Fix the closure of Eq. (19) and test sensitivity: first state the intended boundary conditions and make the source term dimensionally consistent, e.g., replace rhoI/rho by (rhoI/rho)/T_ref or rhoI/rho times an explicit rate as in the underlying model [27]. Then rerun the one-exit ventilation scenario of Section 4.1.2 (umax = 1.4 m/s, C0 = 1.2, uin = 10 m/s) under two contrasting boundary-condition sets: (A) beta = 0 Dirichlet at inflow ventilation ducts and zero-gradient elsewhere; (B) zero-gradient on all boundaries. Compare the percentage of exposed pedestrians and the ordering of 'along airflow' versus 'against airflow' outcomes.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's headline ventilation conclusions all pass through the drift-reaction-diffusion equation for beta, Eq. (19). As stated, that equation is not a well-posed initial-boundary-value problem: it supplies an initial condition, beta(x,0)=0, but no boundary conditions on the computational domain. The inflow and outflow ducts are precisely where boundary conditions on beta matter most, because 'clean air' entering through a ventilation duct and the decay of beta near an exhaust duct control the spatial pockets that drive the reported exposure reductions. An unspecified choice--for example Dirichlet beta=0 at inflow ducts versus homogeneous Neumann conditions everywhere--can change the effective source of contamination entering the room with the ventilation flow, and therefore can change both the magnitude of the exposed fraction and the ranking of airflow directions. The same equation also contains a dimensional inconsistency. In Eqs. (15)-(17), betaI = i0 beta must have units of inverse time, so beta has units of inverse time; then beta_t has units of inverse time squared, while the source term rhoI/rho is dimensionless. If beta is instead interpreted as dimensionless, then the time-derivative, advection, and diffusion terms have units of inverse time while rhoI/rho remains dimensionless. Either way, one term in Eq. (19) has mismatched units, so the quantitative predictions, and therefore the quantitative exposure comparisons, are not unambiguously defined even if the rest of the coupled model is accepted. This concern is more immediately load-bearing than the potential-flow idealization of ventilation: it affects internal model consistency and reproducibility, not only fidelity to real airflow.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a coupled PDE model for crowd flow and epidemic spreading in confined spaces. The crowd component is a second-order Payne-Whitham-style model with an Eikonal equation for the desired walking direction; the epidemic component is a macroscopic SEISV model; infection transmission is represented through a drift-diffusion equation for the infection coefficient, with air flow obtained from a potential-flow solution of Laplace's equation. The authors implement this model with finite volume schemes (Roe for the crowd model and Rusanov for the advection equations, with MUSCL reconstruction and a fast sweeping method for the Eikonal equation) and run numerical experiments for one-exit, two-exit, and corridor geometries, varying ventilation rate and direction, maximum pedestrian speed, pressure coefficient, exit width, and the fraction of masked/vaccinated individuals. The main reported findings are that total exposure decreases with higher ventilation rate, with air flow against pedestrian motion, with higher maximum speed, with higher pressure coefficient, and with masked/vaccinated individuals, with a noted reversal for large crowds near an exit.","tokens_in":18545,"tokens_out":5273,"duration_ms":52752,"significance":"If taken as a deliberately idealized modeling study, the paper is useful: it constructs a complete coupled PDE system, states all parameter values, and performs a broad sweep of control scenarios without fitting parameters to reproduce the exposure percentages. The numerical infrastructure is standard and the parameter choices are mostly traceable to the cited literature. The main value is qualitative hypothesis generation for macro-control measures in evacuating crowds. However, the quantitative exposure percentages and the rankings of ventilation scenarios are not yet fully grounded, because the infection-coefficient equation has an apparent dimensional inconsistency and missing boundary conditions, and because no grid-convergence or sensitivity study is reported. The ventilation conclusions also rest entirely on an inviscid potential-flow idealization of indoor air movement.","major_comments":[{"comment":"Equation (19) is not dimensionally consistent as written. Since beta_I = i0 beta in Eqs. (15)-(17) is an infection rate with units 1/time, beta itself has units 1/time when i0 is dimensionless. Then beta_t, the advection term, and the diffusion term each carry units 1/time^2, while the source term rho_I/rho is dimensionless. If beta is instead interpreted as dimensionless, the differential terms carry 1/time and the source term remains dimensionless. Either way, one term in Eq. (19) has mismatched units. This directly affects the quantitative exposure percentages in Section 4 and the interpretation of i0. Please introduce an explicit rate constant in the source term, state the units of i0, sigma, nu, and beta, and rerun or justify the simulations with a consistent formulation.","section":"Section 2.2, Eq. (19)"},{"comment":"No boundary conditions are given for the infection-coefficient equation (19). The ghost-cell conditions listed in Section 3 apply to the crowd flow model only, and Section 2.2 supplies only beta(x,0)=0. Boundary conditions matter most at the inflow and outflow ducts, where the clean-air pockets and ventilation-direction effects in Figs. 7-12 and 22 are controlled by how beta is transported into and out of the domain. A Dirichlet condition beta=0 at supply ducts versus homogeneous Neumann conditions at walls and exhausts can change both the magnitude of the exposed fraction and the ranking of airflow directions. Please specify the boundary conditions used in the simulations and test the sensitivity of the reported trends to that choice.","section":"Section 3, boundary conditions paragraph"},{"comment":"No grid-convergence or sensitivity study is reported. All one-exit and two-exit runs use Delta x = 0.05 m and Delta t = 2e-3 s, while the corridor runs use Delta x = 0.1 m and Delta t = 5e-3 s, with no refinement study. Many of the reported exposure differences are only a few percentage points, for example the reductions of up to about 6% in Section 4.1.2, so numerical dissipation and dispersion could be of the same order as the effects being claimed. Please add at least one mesh-refinement test for each geometry and report the resulting exposure percentages to confirm the conclusions are not grid-dependent.","section":"Section 4"},{"comment":"The ventilation velocity field U_G is obtained as a potential flow of an inviscid, irrotational, incompressible fluid. This ignores turbulence, crowd-air interactions, and duct geometry, and the spatial structure of U_G is exactly what determines the 'pocket of clean air' near the exit that drives the ventilation-direction conclusions in Figs. 7-8 and Sections 4.1-4.3. The paper should state this limitation more prominently and, for at least one scenario, compare the potential-flow velocity field with a more realistic field, such as a RANS solution or measured data, to show that the qualitative trends are robust.","section":"Section 2.2, Eqs. (20)-(21)"}],"minor_comments":[{"comment":"Equation (20) contains a typo: 'where Psi(x is' should be 'where Psi(x) is'.","section":"Section 2.2, Eq. (20)"},{"comment":"The sentence defining the initial condition states rho0 = rhoS0 + rhoE0 + rhoS0 + rhoV0; the third term should presumably be rhoI0.","section":"Section 2.2, after Eq. (18)"},{"comment":"In the CFL condition (29), 'therm' should be 'term'.","section":"Section 3, CFL condition"},{"comment":"The limiter is named 'van Albaba' but should be 'van Albada'.","section":"Appendix"},{"comment":"The caption says 'Equilibrium speed-density relations for flow (left) and velocity (right)', but both panels appear to be speed-density relations; please clarify what distinguishes the two panels.","section":"Figure 1 caption"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on arXiv:2411.16223. It's a numerical study of a coupled PDE model: a second-order pedestrian flow model, an SEISV epidemic model, and a drift-diffusion equation for the infection coefficient, with ventilation entering as a potential flow. The paper systematically tests ventilation rate and direction, maximum pedestrian speed, a pressure coefficient that controls average spacing, and masked/vaccinated individuals. The most interesting result is that pushing airflow against pedestrian motion helps in small crowds but can hurt in large crowds because the clean-air pocket near the exit is too small. That non-obvious interaction is the paper's real contribution.\n\nThe good: this is a sensible assembly of known components, and the parameter exploration is broader than earlier work by Salam, Bock, and colleagues. The numerics are standard (Roe with MUSCL, Rusanov for transport equations, fast sweeping for the Eikonal), and the qualitative behavior is plausible. No parameters were fitted to generate the exposure numbers; they come from the stated equations.\n\nThe soft spot is load-bearing. Equation (19), which evolves β, is not a well-posed IBVP: the paper gives only the initial condition β(x,0)=0 and no boundary conditions on the ducts or walls. Since ventilation enters exactly through the inflow and outflow boundaries, the choice of BC for β will change the clean-air pockets that drive the exposure comparisons. The stress-test note also checks out: βI = i0 β must have units of 1/time to make sense in the SEISV equations, so β has units of 1/time; then β_t has units of 1/time^2 while the source term ρI/ρ is dimensionless. One term in (19) is dimensionally inconsistent no matter how you read it. This means the quantitative exposure percentages, and possibly some rankings of ventilation direction, are not uniquely determined until the equation is repaired. That's a bigger deal than the potential-flow idealization, which the authors acknowledge as an imitation.\n\nTwo smaller issues: no grid-convergence or sensitivity analysis, and no code or data released. Given the paper is purely numerical, both should be addressed.\n\nNet: the model is worth publishing after revision. The direction-reversal finding is interesting enough that a serious referee should look at it. Ask for a corrected Eq. (19), explicit boundary conditions, and a convergence study. I'd take the qualitative conclusions with a grain of salt until then, but I wouldn't desk-reject it.","headline":"Worth a careful look for the ventilation-direction reversal, but Eq. (19) has a units and boundary-condition problem that undermines the quantitative exposure comparisons until fixed.","tokens_in":19104,"tokens_out":4434,"would_cite":false,"duration_ms":77434,"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":"This paper builds a coupled PDE model of crowd flow, airborne contagion, and ventilation, and uses it to rank macro control measures by how many people become exposed during an evacuation.","keywords":["crowd flow","epidemic spreading","macroscopic models","numerical solution","ventilation control","SEISV model","potential flow","infection coefficient"],"falsifier":"Replace the potential-flow air field with a measured or turbulence-resolving airflow in the same geometries and compare predicted exposed totals; if the direction of airflow that minimizes exposure changes, the ventilation conclusions are artifacts of the ideal airflow assumption.","tokens_in":18036,"feed_emoji":"🦠","tokens_out":8645,"duration_ms":73600,"temperature":0.7,"pith_summary":"This paper asks whether a handful of macroscopic levers—ventilation rate and direction, pedestrians' maximum speed, average spacing, and masked or vaccinated fractions—can be ranked by their effect on disease exposure in confined evacuating crowds. To answer it, the authors couple three PDE components: a second-order crowd-flow model with Eikonal route choice, a Susceptible-Exposed-Infected-Susceptible-Vaccinated contagion model advected by the crowd, and a drift-reaction-diffusion infection coefficient driven by a potential-flow ventilation field. In simulated rooms of one and two exits and a larger corridor, the model predicts that total exposed individuals decrease with higher ventilation rate, airflow against walking direction, higher maximum speed, larger pressure coefficient (that is, larger average interpersonal distance), and inclusion of masked or vaccinated individuals. The same tests also reveal limits: when exits are wide or multiple, spacing matters less, and with narrow exits very high walking speeds can worsen clogging and exposure. If these trends hold, the model gives building operators and safety planners a quantitative way to compare interventions before deploying them.","feed_headline":"Ventilation and crowd speed shift who gets exposed in evacuations","feed_subtitle":"Coupled PDE simulations show fewer exposures with faster airflow, against-flow walking, wider spacing, and masks.","key_machinery":"The carrying object is the infection-coefficient field $\\beta(x,t)$, a transportable quantity that converts infected presence into new exposures: the source term for exposed density is $\\beta_I\\rho_S$. The field is advected by the ventilation velocity $U_G$, diffuses with effective turbulent viscosity $\\sigma$, decays by aerosol settling $\\nu\\beta$, and is replenished proportionally to infected density $\\rho_I/\\rho$. Computing $U_G$ as the gradient of a Laplace potential $\\Psi$ with Neumann duct boundary conditions is what lets ventilation rate and direction enter the model without resolving indoor turbulence. The crowd-flow component supplies both the advection velocity for the epidemic densities and the Eikonal-based desired direction toward exits, so the spatial density profile—not just a uniform mixing assumption—controls where exposure happens.","core_discovery":"This paper's central claim is that epidemic transport in a closed, evacuating space can be described by a single coupled system in which the infection field is not imposed by hand but emerges from crowd motion, ventilation, and infected density. The force of infection is $\\beta_I = i_0\\beta(x,t)$, where $\\beta$ solves a drift-reaction-diffusion equation $\\beta_t + \\nabla\\cdot(\\beta U_G) = \\nabla\\cdot(\\sigma\\nabla\\beta) - \\nu\\beta + \\rho_I/\\rho$, and $U_G = \\nabla\\Psi$ is a steady potential-flow velocity field obtained from Laplace's equation with duct-like boundary conditions. Advecting the SEISV densities with the crowd velocity from a second-order macroscopic flow model closes the loop. In the numerical tests, the predicted exposed total falls when ventilation rate rises, when air moves against pedestrian flow, when maximum speed rises, when pressure coefficient $C_0$ rises (larger average distances), and when masked or vaccinated pedestrians are present; the direction effect reverses in large crowds near an exit, where a clean-air pocket of the size seen in smaller rooms no longer reaches most of the crowd.","pith_inferences":["Beyond the paper, the potential-flow airflow is the linchpin of the ventilation conclusions; replacing it with a turbulence-resolving airflow solver in the same geometries would directly test whether the direction rankings survive real indoor air motion.","Because the model outputs full spatiotemporal density and infection fields, it invites a closed-loop control formulation—ventilation rate, direction, and speed or spacing recommendations updated from measured crowd state—which the paper mentions as a possibility but does not implement.","The simulations fix recovery, latency, and vaccination rates at zero; extending the SEISV equations with nonzero recovery, latency, and vaccination rates would determine whether the same control rankings hold for longer exposures or partially immune populations.","A tracer-gas experiment in a ventilated room with a controlled walking crowd could probe the predicted clean-air pocket at the exit and the reversal of direction preference in large crowds."],"forward_implications":["If the model is right, ventilation rate can be treated as a tunable exposure-reduction knob, but its direction should be chosen per geometry; airflow against walking direction helps small single-exit rooms, while a pattern that sweeps clean air across a large waiting area helps big crowds.","Faster pedestrian movement reduces exposure mainly by shortening evacuation time; the benefit saturates or reverses when a narrow exit turns higher speed into stronger clogging.","Enforcing larger average interpersonal distance (higher $C_0$) lowers exposure and evacuation time in single-exit congestion, yet contributes little when multiple or wide exits already keep densities low.","A fraction of masked or vaccinated pedestrians reduces the exposed total additively with the other levers, bringing the lowest observed exposure in the tests to about 4% of the crowd.","Real-time spatial density and infection-field profiles, rather than room-averaged totals, are the natural signals for deciding when to change ventilation or issue spacing and speed recommendations."],"supporting_citations":[{"why":"Supplies the coupled pedestrian-flow/disease-contagion model structure and the drift-diffusion infection coefficient equation that this paper modifies.","marker":"[27]"},{"why":"Gives the infectivity scale $i_0=0.04$ and the prior mesh-free coupled disease-contagion/crowd-motion formulation.","marker":"[26]"},{"why":"Provides the feedback-control crowd evacuation model that the crowd-flow component is modified from.","marker":"[16]"},{"why":"Supplies the second-order macroscopic pedestrian flow model also used as a basis for the crowd-flow component.","marker":"[22]"},{"why":"Justifies the inviscid, incompressible, irrotational potential-flow assumption from which the ventilation velocity field is computed.","marker":"[17]"},{"why":"Provides the aerosol diffusion characteristics in ventilated rooms that set the turbulent viscosity and settling terms in the infection-coefficient equation.","marker":"[37]"},{"why":"Supplies realistic ventilation rate magnitudes used to calibrate the duct inflow speeds in the simulations.","marker":"[5]"},{"why":"Provides the fast sweeping method used to solve the Eikonal equation for pedestrians' desired walking direction.","marker":"[38]"}],"fun_headline_variants":["Ventilation and crowd speed alter exposure in evacuations","Airflow and spacing cut infection risk in simulated mobs","Faster airflow and wider gaps reduce epidemic spread in crowds","Simulated evacuations map how ventilation and masks curb spread","Coupled PDEs show ventilation and walking speed shape contagion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The ventilation air is treated as a steady, smooth, incompressible potential flow; real indoor airflow is turbulent and altered by the crowd itself, so the exposure reductions attributed to ventilation rate and direction rest on this idealization.","fun_headline_variants_meta":{"raw":{"variants":["Ventilation and crowd speed alter exposure in evacuations","Airflow and spacing cut infection risk in simulated mobs","Faster airflow and wider gaps reduce epidemic spread in crowds","Simulated evacuations map how ventilation and masks curb spread","Coupled PDEs show ventilation and walking speed shape contagion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000192,"raw_usage":{"total_tokens":1388,"prompt_tokens":1027,"completion_tokens":361,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":643,"completion_tokens_details":{"reasoning_tokens":279}},"tokens_in":643,"tokens_out":361,"duration_ms":4569,"temperature":1.0,"reasoning_tokens":279,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:22:56.867400+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Replace the potential-flow air field with a measured or turbulence-resolving airflow in the same geometries and compare predicted exposed totals; if the direction of airflow that minimizes exposure changes, the ventilation conclusions are artifacts of the ideal airflow assumption.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the coupled pedestrian-flow/disease-contagion model structure and the drift-diffusion infection coefficient equation that this paper modifies."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the infectivity scale $i_0=0.04$ and the prior mesh-free coupled disease-contagion/crowd-motion formulation."},{"cited_title":"Kachroo, S.A","cited_arxiv_id":null,"evidence_quote":"Provides the feedback-control crowd evacuation model that the crowd-flow component is modified from."},{"cited_title":"Sundar, and J¨ org Kuhnert","cited_arxiv_id":null,"evidence_quote":"Supplies the second-order macroscopic pedestrian flow model also used as a basis for the crowd-flow component."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Justifies the inviscid, incompressible, irrotational potential-flow assumption from which the ventilation velocity field is computed."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the aerosol diffusion characteristics in ventilated rooms that set the turbulent viscosity and settling terms in the infection-coefficient equation."},{"cited_title":"Ventilation for acceptable indoor air quality, 2015","cited_arxiv_id":null,"evidence_quote":"Supplies realistic ventilation rate magnitudes used to calibrate the duct inflow speeds in the simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the fast sweeping method used to solve the Eikonal equation for pedestrians' desired walking direction."}],"review_version":1}