Animal movement between populations based on family trees - a test case on dugongs in southern Queensland

Abstract

A pedigree is one of the most intuitive concepts in biology, due to our familiarity with family trees for humans and livestock. A pedigree structure, linking parents to their offspring, is a high-resolution snapshot of a population giving insight into the dynamics of genetic and social structure, including: relatedness of individuals, coefficients of co-ancestry, levels of inbreeding, and breeding strategy. For wildlife populations, however, determining a pedigree is challenging. Pedigrees may be elucidated through intensive longitudinal studies of populations small enough that each individual can be identified and then followed throughout its lifetime, but such studies are rare, expensive, or even impossible for some cryptic species. When a species has distinct generational structure and large full-sib groups, genetic sibship or parentage assignment methods can provide significant insight. However, for long-lived species with few offspring and little background into generational structure, pedigree reconstruction requires a different approach. In this research project, I developed and validated a new pedigree reconstruction system PR-genie, based on high resolution genetic and ancillary biological data for wildlife species with complex, multi-generational pedigrees and for when entire populations cannot be sampled. The utility of pedigree reconstruction in answering questions related to social dynamics for cryptic species was then demonstrated, in particular the detection of contemporary movement and breeding events between habitat locations, and analysis of breeding strategy. A genetic mark-recapture population program for a fully marine mammal, the dugong (Dugong dugon), has been in progress in southern Queensland, Australia since 2001, and it is to this population that I applied pedigree reconstruction. The pedigree reconstruction system PR-genie was driven via the Cross-Entropy method and is unique in the way that it uses observed body size class data to infer relative age and thus improve accuracy and efficiency of determining parentage and relatedness. A detailed population simulation system, featuring assortative and inter-generational mating, was developed and used to test the system, and comparisons of the resultant pedigrees were made with an alternate previously published reconstruction system, FRANz. Testing was performed over a broad range of genetic data quality (between 5 and 50 microsatellite loci with average heterozygosity 0.35-0.8), over a range of population sizes, population sampling frequencies, and with varying levels of available ancillary biological data including sex and body size class (as a proxy for relative age). Pedigree reconstruction was demonstrated to be viable with reasonable levels of genetic data (20-25 loci) but improved markedly in accuracy as the quality of genetic data increased, and with the use of sex and body size class data to discriminate breeding individuals. Pedigree reconstruction was applied to genetic microsatellite and ancillary biological data (sex and body size) for 1002 different dugongs across the three main foraging groups found within southern Queensland (Moreton Bay, Hervey Bay - Great Sandy Straits, and Shoalwater Bay). Parent-offspring links were used to infer contemporary movement and breeding events between these locations when the related individuals were found in different locations. These pedigree links indicated significantly more movement than had been possible to observe previously through physical recapture of individuals in multiple locations in southern Queensland. The pedigree data also suggested that male dugongs moved more frequently than female dugongs, and that these migrations mediated genetic interchange between habitats. A suite of simulated populations were then developed under a range of possible breeding strategies for large mammals, and for each of these resultant populations, pedigrees were reconstructed. A range of pedigree statistics (including mean number of offspring per parent, frequency of full sib pairs, estimated relatedness between parent pairs, and proportion of reproductively active adults) was used to construct classifiers of breeding strategy via supervised learning algorithms. These classifiers were applied to reconstructed dugong pedigrees to draw inferences about likely breeding strategies. Mating strategies of monogamy and dominance were ruled out through pedigree data, with promiscuous mating being the most likely breeding strategy of this population of dugongs. Pedigree reconstruction is challenging, but has wide applicability for deciphering movements and social (breeding) events in wildlife populations. The techniques developed here provide a means of using the genetic and ancillary biological data that may be reasonably collected for wildlife populations, including for cryptic species, to draw inferences that would otherwise be out of reach. Pedigree reconstruction can give unprecedented insight into familial relationships among individuals in a wildlife population, and the implications of these relationships, for any challenging longitudinal wildlife study for which individual identity data are available.