On how to exploit a population given by a difference equation with random parameters

Abstract

We consider a model of an exploited homogeneous population given by a difference equation depending on random parameters. In the absence of exploitation, the development of the population is described by the equation $$X(k+1)=f\bigl(X(k)\bigr), \quad k=1,2,\ldots,$$ where $X(k)$ is the population size or the amount of bioresources at time $k,$ $f(x)$ is a real differentiable function defined on $I=[0,a]$ such that $f(I)\subseteq I.$ At moments $k=1,2,\ldots$, a random fraction of the resource $\omega(k)\in\omega\subseteq[0,1]$ is extracted from the population. The harvesting process can be stopped when the share of the harvested resource exceeds a certain value of $u(k)\in[0,1)$ to keep as much of the population as possible. Then the share of the extracted resource will be equal to $\ell(k)=\min (\omega(k),u(k)).$ The average temporary benefit $H_*$ from the extraction of the resource is equal to the limit of the arithmetic mean from the amount of extracted resource $X(k)\ell(k)$ at moments $1,2,\ldots,k$ when $k\to\infty.$ We solve the problem of choosing the control of the harvesting process, in which the value of $H_*$ can be estimated from below with probability one, as large a number as possible. Estimates of the average time benefit depend on the properties of the function $f(x)$, determining the dynamics of the population; these estimates are obtained for three classes of equations with $f(x)$, having certain properties. The results of the work are illustrated, by numerical examples using dynamic programming based on, that the process of population exploitation is a Markov decision process.