The gamma distribution and weighted multimodal gamma distributions as models of population abundance (Q792260): Difference between revisions

The aim of this paper is to justify the use of the gamma distribution as a model for population abundance. Authors show that a fairly general class of stochastic population growth equations has limit distributions of gamma type or is approximable by gamma. Suppose that n(t) is the number of individuals at time t and the growth law is described by a differential equation (1) \(dn/dt=n(g(n)+h(n)z(t))\) where g(n) is the specific growth rate, z(t) is a Gaussian white noise ''process'' and h(n) specifies the density dependence in the effects of the noise. Authors develop an explicit formula for the limiting probability density f(n) of n \[ (2)\quad f(n)=\psi \exp 2\sigma^{-2}(\int n^{- 1}h^{-2}(n)g(n)dn-\omega \log nh(n)) \] where \(\sigma^ 2\) is the variability parameter of the noise, \(\psi\) is a norming constant and \(\omega =\sigma^ 2\) if the Ito integral is used in (1) or \(\omega =\sigma^ 2/2\) if the Stratonovich integral is used. The density f(n) can be approximated by a gamma density \((\alpha^ s/\Gamma(s))n^{s- 1}e^{-\alpha n}\), where \(\alpha =(2/(\sigma^ 2h(\bar n)))(\omega h'(\bar n)-g'(\bar n)/h(\bar n))\), \(s=-(2\bar ng'(\bar n)/(\sigma h(\bar n))^ 2)+1-(2\omega /\sigma^ 2)\) and \(\bar n\) is the equilibrium of (1) in the absence of noise. Authors argue that gamma approximation is more relevant than the normal since it reproduces skewness observed in the field data. Several examples of stochastic growth equations are demonstrated to imply limit distributions approximable by gamma (including the logistic equation, flour-beetle model, fisheries models and species abundance models). In addition to the basic model (1), another model is considered, the noise-free version of which can have multiple stable and unstable equilibria (3) \(dn/dt=n(g(n)-p(n)+h(n)z(t)).\) Here, p(n) is a ''removal rate'' due to predation, harvesting, etc. This model yields, at equilibrium, a weighted gamma distribution which can be multimodal. Several examples are provided, including convex and sigmoidal removal rates.