Abstract
In this paper, we use a series of small world networks to simulate the epidemic spreading in the real world. To make our model more similar to the real world, we employ a parameter $p_{move}$ to denote its moving probability, which corresponds with the general mobility of individuals in realistic social network. The networks keep the same small world properties when they are varying. And the SIRS model is used to simulate the disease propagation. From these simulations we see the moving probability of a small world network has a profound effect on the velocity for the spread of an infection. The quantitative results show that the characteristic time $\tau$ of infection at initial stage decreases exponentially with $p_{move}$.
Introduction
Traditional epidemic modeling often relies on two extremes: "well-mixed" populations, where any individual can interact with any other, or "lattices," where interactions are strictly local. While the "small-world" model introduced by Watts and Strogatz in 1998 provided a more realistic middle ground, it typically assumes a static network structure. However, real-world social networks are inherently dynamic; individuals travel, meet new people, and shift their social circles constantly. This paper addresses how the dynamic rewiring of social connections—driven by human mobility—affects the velocity of a disease outbreak.
Methodology
The researchers developed a dynamic small-world network model by introducing a "moving probability" ($p_{move}$). Starting with a standard Watts-Strogatz ring lattice, $p_{move}$ determines the probability at each time step that a connection between two nodes is reconnected based on a template network. This approach ensures the network maintains its characteristic short paths and high clustering even as specific links change.
Disease progression was simulated using the Susceptible-Infected-Refractory-Susceptible (SIRS) model. In this framework, infected individuals (I) remain contagious for a duration $\tau_I$, after which they enter a refractory state (R) for a duration $\tau_R$ before becoming susceptible (S) again. Numerical simulations were conducted on synthetic networks ranging from $10^4$ to $10^6$ nodes, allowing for precise control over variables like node count ($N$), average neighbors ($K$), and mobility rate ($p_{move}$).
Key Results and the Impact of Mobility
The simulations reveal that the moving probability has a profound effect on the speed of infection spread. As shown in Figure 1, even small increases in $p_{move}$ lead to significantly faster growth in the number of infected individuals ($N_{inf}$) and a higher peak infection level.
In the initial stages, the number of infected individuals grows exponentially according to $n_{inf}(t) \sim \exp(t/\tau)$, where $\tau$ is the characteristic time of infection. A key finding of this study is the exponential relationship between this characteristic time and the mobility of the population.
Analysis of Characteristic Time
The most striking insight is the extreme sensitivity of the system to initial movement. The characteristic time $\tau$ follows an exponential decay curve: $\tau = \tau_0 + \tau_1 e^{-p_{move}/p_{mc}}$.
As illustrated in the figure above, the characteristic moving probability ($p_{mc}$) is remarkably small (approximately 0.0064). This implies that even a minimal degree of social "churn" causes a massive leap in the epidemic's velocity. However, this effect eventually saturates; once $p_{move}$ reaches approximately 0.02, the network is sufficiently shuffled that further increases in mobility do not significantly accelerate the spread.
Discussion and Policy Implications
The quantitative relationship discovered, $\tau = 11.72 + 8.77e^{-p_{move}/0.0064}$, suggests a "cliff-edge" for public health interventions. Because infection speed is so sensitive to low levels of mobility, social distancing measures may only be effective if they restrict human activity to an extremely low degree ($p_{move} \approx 0$).
While the study provides strong numerical evidence for these dynamics, the authors note that an analytical derivation of the exponential relationship remains an area for future work. Additionally, the model assumes uniform mobility across the population, whereas real-world mobility often follows a "long-tailed" distribution where a few individuals travel significantly more than others.
Summary updated