Quasistationarity and extinction for population processes

TL;DR

This paper develops analytical approximations for the quasistationary distribution and expected time to extinction in stochastic population processes using WKB methods, with applications to birth-death models.

Contribution

It introduces a Hamilton-Jacobi PDE approach to approximate quasistationary distributions and extinction times, providing explicit solutions and conditions for their validity.

Findings

Analytical solutions for Hamilton-Jacobi equations in population models.

Conditions for improved approximations of quasistationary distributions.

Explicit approximations for extinction times in multitype birth-death processes.

Abstract

We consider stochastic population processes that are almost surely absorbed at the origin within finite time. Our interest is in the quasistationary distribution, $\boldsymbol{u}$, and the expected time, $\tau$, from quasistationarity to extinction, both of which we study via WKB approximation. This approach involves solving a Hamilton-Jacobi partial differential equation specific to the model. We provide conditions under which analytical solution of the Hamilton-Jacobi equation is possible, and give the solution. This provides a first approximation to both $\boldsymbol{u}$ and $\tau$. We provide further conditions under which a corresponding `transport equation' may be solved, leading to an improved approximation of $\boldsymbol{u}$. For multitype birth and death processes, we then consider an alternative approximation for $\boldsymbol{u}$ that is valid close to the origin, provide conditions under which the elements of this alternative approximation may be found explicitly, and hence derive an improved approximation for $\tau$. We illustrate our results in a number of applications.