A simulation study of bias in two standard mark-recapture estimators when used on populations with differing social structures under different trapping scenarios

Abstract

The estimation of population size is important in studying population dynamics (Kingsland 1995, Turchin 2003), in calculating effective population sizes in studies of evolutionary dynamics (Nunney and Elam 1994, Frankham 1995), in estimating the sex-ratio of populations (Clutton-Brock 1986, Clutton-Brock and Iason 1986), and in monitoring the status of populations which are at risk of extinction (Jones et al. 2013). The estimation of population size in the wild can be challenging as animals may range over vast, inaccessible areas and individuals may not be easily detectable. Social species may present an additional challenge as individuals belonging to different social groups may have varying detectability (Cubaynes et al. 2010), especially if these groups have changing compositions over short time periods. While capture-recapture models have been widely used, and progressively refined, in order to estimate the population sizes of animals in the wild (Seber 1982, Williams et al. 2002, Amstrup et al. 2005), these models do not explicitly consider sociality as a factor that can affect population estimates. Therefore, in this thesis, I examine whether the current statistical methods of population size estimation give unbiased estimates for populations exhibiting different kinds of social structures. Using individual-based simulations, I also explore the effect of population densities, trap densities (sampling intensity), sampling scale, and trap spatial arrangement or distribution on the efficacy of two commonly used markrecapture estimators, POPAN and Robust Design with heterogeneity.