Ergodic property for Galton-Watson processes in which individuals have variable life times
This paper is concerned with an extended Galton-Watson process so as to allow individuals to live and reproduce for more than one unit time. We assume that each individual can live $k$ seasons (time-units) with probability $h_k$, and produce $m$ offspring with probability $p_m$ during each season. These can be seen as Galton-Watson processes with countably infinitely many types in which particles of type $i$ may only have offspring of type $i+1$ and type $1$. Let $\textbf{M}$ be its mean progeny matrix and $γ$ be the convergence radius of the power series $\sum_{k\geq 0}r^k(\textbf{M}^k)_{ij}$. We first derive formula of calculating $γ$ and show that $γ$, in supercritical case, is actually the extinction probability of a Galton-Watson process. Next, we give clear criteria for $\textbf{M}$ to be $γ$-transient, $γ$-positive and $γ$-null recurrent from which the ergodic property of the process is discussed. The criteria for $γ$ and $γ$-recurrence of $\textbf{M}$ rely on the properties of lifetime distribution which are easier to be verified than current results. Finally, we show the asymptotic behavior of the total population size of each type of individuals under certain conditions which illustrates the evolution of Galton-Watson process in which individuals have variable lifetimes.