Demographic inference in a spatially-explicit ecological model from genomic data: a proof of concept for the Mojave Desert Tortoise

Abstract

In this paper, we study the general problem of extracting information from spatially explicit genomic data to inform inference of ecologically and geographically realistic population models. We describe methods and apply them to simulations motivated by the demography of the Mojave desert tortoise (Gopherus agassizii). The tortoise is an example of a long-lived, threatened species for which we have an excellent understanding of range, habitat preference, and certain aspects of demography, but inadequate information on other life history components that are important for conservation management. We use an individual-based model on a discretized geographic landscape with overlapping generations and age and sex-specific dispersal, fecundity, and mortality to develop and test a method that uses genomic data to infer demographic parameters. We do this by seeking parameters that best match a set of spatial statistics of genomes, which we introduce and discuss. We find that for inferring only overall population density and mean migration distance, a simple statistical learning method performs well using simulated training data, inferring parameters to within 10% accuracy. In the process, we introduce spatial analogues of common population genetics statistics, and discuss how and why they are expected to contain signal about the geography of population dynamics that are key for ecological modeling generally and conservation of endangered taxa.