Simulating City-level Airborne Infectious Diseases

With the exponential growth in the world population and the constant increase in human mobility, the danger of outbreaks of epidemics is rising. Especially in high density urban areas such as public transport and transfer points, where people come in close proximity of each other, we observe a dramatic increase in the transmission of airborne viruses and related pathogens. It is essential to have a good understanding of the `transmission highways’ in such areas, in order to prevent or to predict the spreading of infectious diseases. The approach we take is to combine as much information as is possible, from all relevant sources and integrate this in a simulation environment that allows for scenario testing and decision support. In this paper we lay out a novel approach to study Urban Airborne Disease spreading by combining traffic information, with geo-spatial data, infection dynamics and spreading characteristics.

💡 Research Summary

The paper presents an integrated simulation framework for modeling the spread of airborne infectious diseases at the city scale, combining Geographic Information Systems (GIS) with Multi‑Agent Systems (MAS). Recognizing the heightened risk of epidemics in densely populated urban environments—exemplified by the 2009 H1N1 outbreak in China—the authors argue that a detailed representation of urban infrastructure, human mobility, and behavioral patterns is essential for accurate epidemic forecasting and control.

The authors first construct a hierarchical city model. The city is divided into six region types (housing, office, school, university, medical, recreational), each further partitioned into sub‑locations (housing SL, office SL, classroom SL, patient SL, recreational SL). This spatial granularity enables distinct indoor and outdoor transmission probabilities, reflecting the reduced infectivity of outdoor air due to sunlight, wind, and ventilation.

Two transport networks are then built: a road network (RN) and a public‑transport network (PTN). Both are represented as graphs with nodes (intersections or stops) and edges (road segments or line sections). The RN captures vehicle traffic, while the PTN models bus, tram, and metro lines, including stop locations and line crossings.

Mobility routing is handled with a two‑tier approach. For trips shorter than 3 km, agents travel along a straight line between origin and destination. Longer trips are routed on the RN using Dijkstra’s shortest‑path algorithm, with the option to substitute alternative feasible paths to model traffic avoidance. Public‑transport routing employs a breadth‑first search that iteratively expands a radius around the origin and destination to locate nearby stops, then follows line connectivity (directly reachable stops, DR) to construct a feasible itinerary. The radius parameter is tunable, allowing the model to reflect varying willingness to walk or transfer distances.

The synthetic population is generated using MAS principles. Each agent is assigned demographic (age, gender), epidemiological (susceptibility, immunity), and socioeconomic attributes (social status, household composition). Housing and office sub‑locations are allocated based on statistical distributions of household formation and commuting distances. Agents follow daily agendas derived from empirical activity patterns: staying at home, working in offices, attending school or university, visiting recreational venues, and seeking medical care when infected. Activity durations and variances are taken from prior studies (e.g., Valle et al., 2006). During each activity, agents interact with others present in the same sub‑location, and infection events are probabilistically determined using indoor versus outdoor transmission rates.

The simulation proceeds in discrete time steps, updating agents’ locations, contacts, infection states (susceptible, infected, infectious, treated, recovered), and health outcomes. Policy scenarios—such as restricting certain regions, reducing public‑transport frequency, or enforcing mask usage—can be introduced, and the model quantifies their impact on the spatial spread and total case count. By mapping high‑frequency contact pathways, the framework identifies “transmission highways” within the urban fabric, offering decision‑makers actionable insights for targeted interventions.

The paper’s contributions lie in its comprehensive integration of GIS‑based spatial data, realistic transport routing, and behaviorally rich agent modeling. This allows simultaneous consideration of physical proximity, movement patterns, and social behavior—elements often treated separately in prior epidemic models. However, limitations are acknowledged: the current implementation relies on static mobility data and fixed epidemiological parameters, lacking real‑time GPS or traffic flow inputs that could capture dynamic changes in crowding or compliance. Moreover, the transmission parameters are not calibrated to specific pathogens beyond generic airborne assumptions.

Future work is suggested to incorporate live mobility feeds (mobile phone location data, traffic sensors) and environmental measurements (air quality, temperature, humidity) to enable adaptive parameter updates. Extending the model to accommodate pathogen‑specific characteristics (e.g., varying incubation periods, asymptomatic transmission) would broaden its applicability. Overall, the study provides a solid methodological foundation for city‑level airborne disease simulation and highlights pathways for enhancing realism and policy relevance.