Public Transport & Epidemic Coupling

• Public Transport and Epidemic Coupling is the study of how transit systems facilitate disease spread by creating dense, recurrent contacts through both travel and activity sites.
• Mathematical frameworks like the Trans-SEIR model capture detailed agent mobility states, revealing that on-vehicle contagion and location-based mixing can amplify transmission and synchronize outbreaks.
• Empirical research demonstrates that targeted transit interventions—such as capacity controls and enhanced screening—can substantially reduce epidemic peaks and overall infection rates.

Public transport and epidemic coupling refers to the explicit, mathematically resolved interactions between infectious disease transmission dynamics and urban mobility, especially as shaped by transit systems. State-of-the-art research demonstrates that public transport networks—by inducing dense, spatially extended, and temporally recurrent contacts—actively synchronize, amplify, and redistribute epidemic risks across urban landscapes. Models now resolve contagion events both during travel (in vehicles or stations) and at activity sites, quantify intervention impacts (e.g., capacity constraints, targeted screening, and closure policies), and expose critical threshold phenomena arising from mobility structure.

1. Mathematical Frameworks for Coupled Transport–Epidemic Dynamics

The canonical approach decomposes the population into mobility and infection states, with transitions governed by day-level stochastic or deterministic processes. The Trans-SEIR model provides a detailed account: each agent occupies one of four disease states (S, E, I, R) and one of four mobility states (“home,” “travel to activity,” “activity location,” “return travel”). Movements are parameterized by departure rates, mode-specific travel times, and destination matrices calibrated from travel surveys. Upon achieving mobility equilibrium, the SEIR process overlays epidemic transitions within each pool, tracking distinct contributions from activity-based (site) and travel-based (in-vehicle) contacts (Qian et al., 2020).

The governing ODEs for each pool include contributions from inflow/outflow (mobility), on-site infection, and in-vehicle contagion, and yield a high-dimensional, block-structured system (e.g., 2,760 ODEs for 15 New York City boroughs with 4×4 pools each).

2. Transmission Pathways: Activity-Based and Travel-Based Contagion

In this coupled paradigm, two infection mechanisms are dominant:

• Activity-based contagion: Standard mass-action mixing at sites of collective activity (e.g., workplaces, schools), formalized via per-contact site-specific transmission parameters $\beta^A_j$ and contact rates $\kappa_{ij}$. The force of infection for pool $S_{ij}$ at site $j$ is $$f_{ij}(S, I) = \beta^A_j \kappa_{ij} \frac{S_{ij}(I_j + \sum_k I_{kj})}{N^p_j}$$, with $N^p_j$ the present population at $j$.
• Travel contagion: On-vehicle mixing is quantified by mode-specific transmission rates $\beta^T_d$, contact matrices $\kappa_{(k\ell),(ij)}^d$, and mode shares $c^d_{ij}$. The infection increment is $\kappa_{ij}$0.

This explicit separation clarifies both analytic and control properties: in-network disease reproduction number $\kappa_{ij}$1 decomposes into activity and travel components. If travel contagion is negligible ($\kappa_{ij}$2), classical spatial SEIR thresholds are recovered. When travel terms are non-negligible, ignoring them underestimates $\kappa_{ij}$3 and delays epidemic peaks by weeks (Qian et al., 2020, Hajdu et al., 27 Jan 2025).

3. Thresholds, Synchronization, and Critical Mobility

Several research lines identify sharp coupling-induced thresholds and synchronization patterns:

• Critical mobility levels: Agent-based studies show a well-defined radius-of-gyration $\kappa_{ij}$4 below which outbreaks die out and above which major waves ensue. At fixed immunity, the critical value $\kappa_{ij}$5 is parameterized analytically, and effective transmission $\kappa_{ij}$6 grows linearly with $\kappa_{ij}$7 (López et al., 2024).
• Network-induced synchronization: Travel contagion compresses temporal asynchrony across urban zones, exponentially reducing variance in local infection growth rates and promoting peak-aligned outbreaks at city scale (Qian et al., 2020).
• Occupancy-driven $\kappa_{ij}$8: Bus simulations show $\kappa_{ij}$9 varies almost linearly with load fraction, with a threshold load of $S_{ij}$0 for epidemic subcriticality—a value robust to other contact-network manipulations (Hajdu et al., 27 Jan 2025).
• Diversity-lowered thresholds: When the variance (diversity) in individual public-transport use is increased, the epidemic threshold is lowered. A population with identical ride frequencies is more epidemic-resilient than one with a wide spread of “supercommuters” and near-non-users, even at the same system-wide average utilization (Zhou et al., 2012).

4. Empirical and Structural Insights from Case Studies

Calibrated applications have revealed spatial and network-level nuances:

New York City COVID-19:

• 29% of infections during the first surge (late March) arose during travel, with transit exposure accounting for up to 40% in northern Manhattan and only 15% on Staten Island. High-volume in-flows into East Manhattan were the most critical travel–contagion sources.
• Optimized entrance screening systems prioritizing high-flow trips into hub boroughs produced significant reductions in $S_{ij}$1 with modest (<5% of total riders) daily screening capacity (Qian et al., 2020).

China's intercity epidemic diffusion:

• Multiplex, bi-partite models across 347 cities revealed that airlines preferentially seed long-distance introductions but are minor sources of in-flight transmission. Buses and regional rail concentrate within-province and short-range spillover, enhanced by cross-infection at intermediate stops (Li, 2020).

Network-based risk profiling:

• Transfer/community network approaches leveraging smartcard data can rank vehicle trips/routes by cumulative infection probability, consistently identifying urban-core and peak-hour trunk lines as highest risk. Interventions targeting these high-risk “super-spreader” routes achieve a higher marginal return (Hajdu et al., 2018, Shoghri et al., 2020).

5. Interventions: Capacity, Demand, and Control Policy

Intervention design benefits from explicit coupling models:

• Capacity restrictions: Imposing vehicle caps alone is sublinear in efficacy—global attack rates decrease at best 3–5% per 10% capacity cut, while the number of stranded passengers rises sharply. A mixed strategy, combining moderate capacity trims and demand management (work-from-home, staggered hours), best balances output and risk, yielding global infection reductions of over 30% with <5% service loss (Hajdu et al., 27 Jan 2025, Hiermann et al., 9 Nov 2025).
• Critical-mobility control: Analytic feedback schemes maintaining population-level mobility just under the empirically measured $S_{ij}$2 suppress incidence with negligible oscillations, outperforming naive “traffic-light” policies that trigger recurrent epidemic waves due to lag effects (López et al., 2024).
• High-risk zone and group targeting: High “super-spreader” and “super-susceptible” transport nodes—typically suburban bus interchanges and some railway midpoints, not always CBD rail stations—should receive enhanced screening, ventilation, load management, and communication. Occupationally distinct groups (e.g., healthcare workers) may function as in-vehicle superspreaders, requiring prioritized mitigation (Chin et al., 2020, Ponte et al., 2021).
• Complexity-aware reopening: Integer/flow-optimization models coupled to simplified SEIR infection endpoints enable globally optimal reopening (lines, stations, timetables) under pandemic constraints, quantifying the trade-off between network accessibility and aggregate infection risk (Huang et al., 2021, Hiermann et al., 9 Nov 2025).

6. Limitations, Caveats, and Data Corrections

Biases arise in models relying solely on public-transit data to infer citywide movement. Excluding private and nonmotorized trips substantially underestimates spatial-temporal epidemic coupling: simulated peaks may be up to 10 days late and ≈10% lower. Corrective approaches include fusing mobile-phone trace data, statistical adjustment of trip matrices, or synthetic modeling with flexible link-distance distributions (Du et al., 2018).

Common simplifying assumptions—static network topology, homogeneous mixing within vehicles/sites, and homogeneous transmission rates—have quantitative but not qualitative impacts. Extensions that resolve finer temporal co-location structure, agent-type heterogeneity, and multi-modal networks are an active area (Shan et al., 2011, Boscheri et al., 2020).

7. Summary Table: Key Models and Features

Model type	Epidemic process	Mobility/Transport representation	Notable findings
Trans-SEIR	SEIR (ODE)	4-state agent mobility with activity+travel	Travel contagion amplifies R₀, synchronizes outbreaks, exposes zone-level gradients (Qian et al., 2020)
Agent-based critical mobility	SEIR (ABM)	Daily trips across spatial zones	Sharp critical mobility thresholds, robust control via $S_{ij}$3 feedback (López et al., 2024)
Network contact/risk ranking	SI/SIR (network)	Time–resolved co-ride, transfer, community graphs, OD flows	Peak-hour and central routes highest risk (Hajdu et al., 2018, Shoghri et al., 2020)
Multilayer metapopulation	SIR (patch ODE)	Intercity trips, multiplexed modes	Route/mode-specific seeding patterns, role of hubs in diffusive spread (Li, 2020)
Space–time/flow optimization	SIR/SEIR (LP+ODE)	Time-expanded transit graphs, integer optimization	Data-driven risk-accessibility trade-off, timetabling for epidemic control (Huang et al., 2021, Hiermann et al., 9 Nov 2025)

• (Qian et al., 2020) "Modeling the spread of infectious disease in urban areas with travel contagion"
• (López et al., 2024) "Critical mobility in policy making for epidemic containment"
• (Hajdu et al., 27 Jan 2025) "Epidemics on the Move: How Public Transport Demand and Capacity Shape Disease Spread"
• (Hajdu et al., 2018) "Discovering the hidden community structure of public transportation networks"
• (Boscheri et al., 2020) "Modeling and simulating the spatial spread of an epidemic through multiscale kinetic transport equations"
• (Li, 2020) "Simulating the Spread of Epidemics in China on the Multi-layer Transportation Network: Beyond the Coronavirus in Wuhan"
• (Du et al., 2018) "Detecting the impact of public transit on the transmission of epidemics"
• (Hiermann et al., 9 Nov 2025) "Public Transport Under Epidemic Conditions: Nonlinear Trade-Offs Between Risk and Accessibility"
• (Zhou et al., 2012) "Epidemic spreading induced by diversity of agents' mobility"
• (Ponte et al., 2021) "Tracing contacts to evaluate the transmission of COVID-19 from highly exposed individuals in public transportation"
• (Chin et al., 2020) "Spatial super-spreaders and super-susceptibles in human movement networks"
• (Shan et al., 2011) "Simulating City-level Airborne Infectious Diseases"
• (Huang et al., 2021) "Optimizing timetable and network reopen plans for public transportation networks during a COVID19-like pandemic"
• (Du et al., 2018) "Urban contact structures for epidemic simulations: Correcting biases in data-driven approaches"
• (Shoghri et al., 2020) "Identifying highly influential travellers for spreading disease on a public transport system"

These works collectively reveal the inherently nonlinear, threshold-driven, and structure-sensitive nature of public transport and epidemic coupling, and provide the theoretical and practical basis for epidemic-resilient, data-optimized mobility networks.