Endogenous Quasicycles and Stochastic Coherence in a Closed Endemic Model

TL;DR

This paper investigates how demographic fluctuations induce sustained quasicycles in a stochastic SIRS model, revealing that stochastic coherence arises from the interplay between eigenvalue properties and population size, driven by detailed balance violation.

Contribution

It demonstrates that stochastic coherence in endemic models can be predicted from deterministic eigenvalue analysis, linking non-normality and detailed balance violation to oscillation regularity.

Findings

Endemic quasicycles are driven by demographic noise and eigenvalue properties.

Maximum coherence occurs at an optimal population size due to non-monotonic eigenvalue ratios.

Violation of detailed balance enhances fluctuations and influences oscillation regularity.

Abstract

We study the role of demographic fluctuations in typical endemics as exemplified by the stochastic SIRS model. The birth-death master equation of the model is simulated using exact numerics and analysed within the linear noise approximation. The endemic fixed point is unstable to internal demographic noise, and leads to sustained oscillations. This is ensured when the eigenvalues ($\lambda$) of the linearised drift matrix are complex, which in turn, is possible only if detailed balance is violated. In the oscillatory state, the phases decorrelate asymptotically, distinguishing such oscillations from those produced by external periodic forcing. These so-called quasicycles are of sufficient strength to be detected reliably only when the ratio $|Im(\lambda)/Re(\lambda)|$ is of order unity. The coherence or regularity of these oscillations show a maximum as a function of population size, an effect known variously as stochastic coherence or coherence resonance. We find that stochastic coherence can be simply understood as resulting from a non-monotonic variation of $|Im(\lambda)/Re(\lambda)|$ with population size. Thus, within the linear noise approximation, stochastic coherence can be predicted from a purely deterministic analysis. The non-normality of the linearised drift matrix, associated with the violation of detailed balance, leads to enhanced fluctuations in the population amplitudes.