Utility of Particle Swarm Optimization in Statistical Population Reconstruction

Statistical population reconstruction models based on maximum likelihood and minimum chi-square objective functions provide a robust versatile approach to estimating the demographic dynamics of harvested populations wildlife. These employ numerical optimization techniques determine which set model parameters best describes observed age-at-harvest, catch-effort, other auxiliary field data. Although numerous methods have been used in past, benefits using particle swarm (PSO) yet be explored. Using North American river otter (Lontra canadensis) Indiana as case study, we investigated performance optimization, spectral projected gradient (SPG), Nelder–Mead, Broyden–Fletcher–Goldfarb–Shanno (BFGS) methods. We Monte Carlo studies simulate under wide range conditions compare relative each four found that consistently significantly improved stability precision when compared with may statistical reconstruction. Given these are frequently guide management decisions harvest limits, encourage agencies adopt this more precise method corresponding abundance. results illustrate caution against relying highly dependent initial conditions, reinforce need ensure convergence global rather than local maximum.