I present a spatially-explicit discriminative modeling approach to infer the location of origin of domesticated plant and animal species from genetic and morphometric variation data. The model is based on the expected monotonic reduction in diversity with geographic distance from region of origin. Such a pattern is expected because as a population expands in space, variation is sampled on the wavefront of expansion, leading to a loss of diversity. My approach performs a search geographic space to identify the region where this correlation is maximized. I account for sparse and uneven sampling, and the possibility of high homozygosity through selfing in plant and animal species, by implementing a spatial kernel. I include a permutation test in order to assign significance for inference of location of origin.
The method has been applied to various species such as broomcorn millet (Panicum miliaceum) microsatellite data to infer that the crop spread from northeast China. We also apply our approach to morphological variation data of the Polynesian rat (Rattus exulans) to investigate the origin of expansion.