With the recent, high-profile public discourse surrounding the potential for virulent outbreaks—be they from an outbreak of Avian Flu or an incident of bioterrorism—epidemiological modeling has been a vital tool used to assess the threat of an epidemic facing the nation. Epidemiological models often have two key components, a method of modeling the spread of an infection within a population and the spread of an infection spatially, over geographic distances, to different population centers.

Modeling the spread of an infection within a population usually relies on fairly simple, but robust, model involving the movement of individuals in and out of various subpopulations (susceptible individuals, infected individuals, and recovered individuals, etc.) dependent on various factors such as the virulence of the disease and the recovery rate/recovery time. However, modeling the spread of an infection over vast spatial dimensions presents a much more daunting task as it requires deriving a viable model of the human travel network and travel behavior.

The simplest model of virulent diffusion geographically is dependent solely on distance, in which areas closer to an infected region are more likely to also be infected than regions farther away. However, diffusive models based solely on geographic distance do not comprehensively capture the phenomena of modern transportation, particularly air travel, in which infected individual could potentially travel large distances in a short time. To correct this, one could introduce the factor of population into the diffusion equation, as transportation hubs tend to be located in areas of high population. Again though, this is still a generalization that may not provide the level of real-world fidelity necessary.

So how does one solve the problem of generating an empirically based model of human travel?

One could certainly attempt to compile a tremendous dataset of all air and train travel, using data from airlines and rail services. However, travel on the highways via personal vehicles is a far less well-documented phenomenon and certainly cannot be ignored.

Surprisingly, a group of German and American scientists turned to what could be considered an internet novelty in their search for data. Dirk Brockmann, et al., used data supplied by the creator of the website Wheresgeorge.com in order to general a diffusion model for human travel in the United States. The site allows users to enter the serial number of a dollar bill, along with the user’s location, and will present all the prior locations the bill was logged at. The team interpreted the trajectories of 464,670 dollar bills in 1,033,095 reports of bill locations as corresponding to human travel. The result of their work with the data provided by WheresGeorge was published in the article The scaling laws of human travel, which appeared in the journal Nature.