[Paper Review] Dynamical Theory and Cellular Automata Simulations of Pandemic Spread: Understanding Different Temporal Patterns of Infections

Saumyak Mukherjee, Sayantan Mondal+1 more|arXiv (Cornell University)|Apr 30, 2020

Mathematical and Theoretical Epidemiology and Ecology Models4 citations

TL;DR

This paper proposes a generalized, non-local SIR model with time-dependent transmission rates and population heterogeneity (high/low risk) to better capture the complex temporal dynamics of pandemic spread, particularly for COVID-19. Using both numerical solutions of coupled differential equations and stochastic cellular automata simulations, it demonstrates that time-varying infection rates and strict home quarantine are critical for accurately modeling infection peaks and controlling outbreaks.

ABSTRACT

Here we propose and implement a generalized mathematical model to find the time evolution of population in infectious diseases and apply the model to study the recent COVID-19 pandemic. Our model at the core is a non-local generalization of the widely used Kermack-McKendrick(KM) model where the susceptible(S) population evolves into two other categories, namely infectives(I) and removed(R). This is the well-known SIR model in which we further divide both S and I into high and low risk categories. We first formulate a set of non-local dynamical equations for the time evolution of distinct population distributions under this categorization in an attempt to describe the general scenario of infectious disease progression. We then solve the non-linear coupled differential equations-(i) numerically by the method of propagation, and (ii) a more flexible and versatile cellular automata (CA) simulation which provides a coarse-grained description of the generalized non-local model. In order to account for multiple factors such as role of spreaders before containment, we introduce a time dependent rate which appears to be essential to explain the sudden spikes before the plateau observed in many cases (for example like China). We demonstrate how this generalized approach allows us to handle the effects of (i) time-dependence of the rate-constants of spread, (ii) different population density, (iii) the age ratio, (iv) quarantine, (v) lockdown, and (vi) social distancing. Our study allows us to make certain predictions regarding the nature of spread with respect to several external parameters, treated as control variables. Analysis of the model clearly shows that due to the strong heterogeneity in the epidemic process originating from the distribution of initial infectives, the theory must be local in character but at the same time connect to a global perspective.

Motivation & Objective

• To address the limitations of the classical SIR model in capturing real-world pandemic dynamics, especially for heterogeneous and non-homogeneous populations.
• To investigate how time-dependent transmission rates, population density, age distribution, and control measures like lockdown and social distancing affect infection spread.
• To develop a more flexible and realistic modeling framework that integrates demographic heterogeneity and non-local transmission effects.
• To evaluate the effectiveness of public health interventions such as home quarantine and social distancing using both dynamical equations and stochastic simulations.
• To provide a predictive tool for policymakers by linking control parameters (e.g., lockdown timing, quarantine compliance) to observable outcomes like peak infection timing and magnitude.

Proposed method

• Proposes a non-local generalization of the Kermack-McKendrick SIR model, dividing susceptible (S) and infected (I) populations into high- and low-risk subgroups.
• Develops a system of non-linear coupled differential equations for the time evolution of S, I, and R (cured + dead) populations with time-dependent infection rate $k_{S\rightarrow I}(t)$.
• Solves the differential equations numerically using a propagation method to simulate the temporal evolution of population fractions.
• Implements a stochastic cellular automata (CA) model that simulates spatio-temporal spread on a grid, incorporating time-dependent transmission probabilities $P_{Tr}(t)$.
• Introduces control parameters such as quarantine compliance, social distancing, and lockdown timing as probabilistic rules in the CA framework.
• Uses conservation laws across sub-populations to ensure total population remains constant, reflecting demographic constraints in the model.

Experimental results

Research questions

• RQ1 How does the inclusion of time-dependent transmission rates affect the shape and timing of infection curves in pandemic models?
• RQ2 To what extent do demographic heterogeneities—such as age distribution, initial population density, and risk group distribution—influence the predicted spread of an infectious disease?
• RQ3 How effective are home quarantine and social distancing in shifting or reducing the peak of infection, according to both dynamical and CA-based simulations?
• RQ4 Can a non-local, generalized SIR model better capture real-world pandemic patterns such as sudden spikes before plateauing, as observed in countries like China?
• RQ5 What role does the incubation period play in delaying detection and increasing transmission risk from asymptomatic carriers?

Key findings

• The time evolution of infection fractions is strongly dependent on time-varying transmission rates, with higher initial rates due to lack of awareness, followed by a decline as public health measures are adopted.
• Cellular automata simulations show that while social distancing reduces transmission risk per outing, it has minimal impact on shifting the peak of infection, unlike effective home quarantine.
• Home quarantine is identified as the most effective intervention, with the fraction of quarantined individuals having a dominant influence on the timing and height of the infection peak.
• Longer incubation periods increase the risk of undetected transmission, as asymptomatic individuals remain infectious longer and are less likely to be isolated.
• The model’s non-linear structure amplifies the effects of initial population heterogeneity, particularly the distribution of initial infectives, making accurate prediction highly sensitive to initial conditions.
• The combination of dynamical equations and CA simulations provides a robust framework that captures both macroscopic trends and microscopic spatial dynamics of disease spread.