Disentangling the relative merits and disadvantages of parentage analysis and assignment tests for inferring population connectivity

Accurately estimating patterns of population connectivity in marine systems remains an elusive goal. Current genetic approaches have focused on assigning individuals back to their natal populations using one of two methods: parentage analyses and assignment tests. Each of these approaches has their relative merits and weaknesses. Here, we illustrate these tradeoffs using a forward-time agent-based model that incorporates relevant natural history and physical oceanography for 135 Kellet’s whelk (Kelletia kelletii) populations from Southern California. Like most marine organisms, Kellet’s whelks live in large metapopulations where local populations are connected by dispersive larvae. For estimating population connectivity, we found parentage analyses to be relatively insensitive to the amount of genetic differentiation among local populations, but highly sensitive to the proportion of the metapopulation sampled. Assignment tests, on the other hand, were relatively insensitive to the proportion of the metapopulation sampled, but highly sensitive to the amount of genetic differentiation found among local populations. Comparisons between the true connectivity matrices (generated by using the true origin of all sampled individuals) and those obtained via parentage analyses and assignment tests reveal that neither approach can explain more than 26% of the variation in true connectivity. Furthermore, even with perfect assignment of all sampled individuals, sampling error alone can introduce noise into the estimated population connectivity matrix. Future work should aim to improve the number of correct assignments without the expense of additional incorrect assignments, perhaps by using dispersal information gleaned from related individuals as priors in a Bayesian framework. These analyses dispel a number of common misconceptions in the field and highlight areas for both future research and methodological improvements.