Numb3rs S1E3: Spatial SIR Model

Season 1, Episode 3 of Numb3rs features an investigation into the source of an unexplained outbreak. After the FBI confirms that the disease is a lab-engineered variant, they suspect it was deliberately released by an individual. To build their case, they need to locate the initial release site. To solve this problem, Charlie applies the SIR Model to trace where the epidemic first emerged. This post explores what the SIR Model is, whether it can be inverted to retroactively identify the outbreak’s origin, and how spatial considerations refine such analyses.

The SIR Model

The SIR Model¹ divides a population into three compartments: S (Susceptible), I (Infected), and R (Recovered), enabling us to model disease dynamics. Their interactions are governed by the system of differential equations:

\begin{align} \frac{dS}{dt} &= - \beta S I, \newline \frac{dI}{dt} &= \beta S I - \gamma I, \newline \frac{dR}{dt} &= \gamma I \end{align}

This model captures two key mechanisms: (1) new infections arise from contact between infected and susceptible individuals, and (2) infected individuals recover or perish, moving them out of the infectious state. While the basic SIR framework is simple, it has spawned numerous extensions— SEIR models that account for an exposed period E, SEIQR models incorporating quarantine Q, and many others. Despite its simplicity, the SIR model has become foundational to modern epidemiological research on disease transmission rates and outbreak forecasting.

Limitations of the SIR Model

However, the SIR model alone cannot account for the kind of source-tracing shown in the episode. The standard SIR model assumes a complete graph: every person has a potential to contact every other person. It ignores geography and realistic human movement patterns entirely.

In reality, contact networks are sparse and depend heavily on individual mobility and geographic proximity. Two people far apart have almost zero chance of contact, and even nearby residents won’t transmit unless both venture outside. Moreover, the presence of hubs—buses, subway stations, shopping centers—where crowds gather dramatically alters the connectivity structure of the network. The Spatial SIR Model extends the classical SIR framework to incorporate such geographic information.

Some approaches focus on population density in space and model diffusion alongside compartmental dynamics². Others account for human mobility to analyze how social distancing affects epidemic control³.

Patient Zero / The Inverse Problem

Even with spatial extensions, Spatial SIR Models are designed for forward modeling—predicting how disease spreads from a known source. Inverting the model to find the source is substantially harder. A common approach is to posit various hypotheses for Patient Zero and the outbreak location, simulate the epidemic forward under each scenario, and identify the scenario whose predicted current state best matches observed data⁴. Computational methods include Monte Carlo sampling or learning-based inference via machine learning⁵.

Real-World Methods

In practice, source reconstruction leverages far more information than epidemiological models alone. For instance, by sequencing the pathogen’s genome, one can construct a mutation phylogeny; combining this with infection timelines and patient movement histories allows investigators to infer transmission order and trace back to an origin. While labor-intensive and impractical for large outbreaks, this approach is feasible for early-stage investigations with few cases—precisely the scenario depicted in the episode. In fact, the show’s revelation that the pathogen matches a known lab variant suggests the investigators employed sequence-based reasoning, aligning with real epidemiological detective work.

References

1. Kermack WO, McKendrick AG (1927). “A Contribution to the Mathematical Theory of Epidemics”. Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character. 115 (772): 700–721. ↩

2. V. Capasso, “Global solution for a diffusive nonlinear deterministic epidemic model,” SIAM Journal on Applied Mathematics, vol. 35, no. 2, pp. 274–284, 1978. ↩

3. Bisin, Alberto and Moro, Andrea, “JUE insight: Learning epidemiology by doing: The empirical implications of a Spatial-SIR model with behavioral responses”, Journal of Urban Economics, vol. 127, 2022. ↩

4. Antulov-Fantulin, Nino and Lančić, Alen and Smuc, Tomislav and Stefančić, Hrvoje and Šikić, Mile, “Identification of Patient Zero in Static and Temporal Networks: Robustness and Limitations”, Phys. Rev. Lett., vol. 114, no. 24, 2015. ↩

5. Sterchi, M., Hilfiker, L., Grütter, R. et al. Active querying approach to epidemic source detection on contact networks. Sci Rep 13, 11363 (2023). ↩