4.1. Current potential distribution ranges of Kersting’s groundnut and genetic groups
We used two approaches to model KG distributions across agroclimatic zones of four West African countries: a traditional species SDM using a species wide cultivated areas, and intraspecific models based on genetic information. We found that decomposing a species into intraspecific genetic groups increases our understanding of potential distribution of KG, which is consistent with results from the previous few studies (Gotelli & Stanton-Geddes 2015; Ikeda et al. 2016; Marcer et al. 2016; Oney et al. 2013). The SDMs for Pop1 performed more accurately in estimating potential cultivation areas of KG rather than the genetic Pop2, which showed lower prediction accuracy. The Pop2 is found in the three agroclimatic zones of all countries; it displays a wide distribution range like the whole-species. Indeed, Allouche et al. (2006) reported that the values of predictive accuracy tend to be lower when species are more common within the distribution ranges. Our results reinforce the view that it is necessary in SDMs evaluation, to consider not only the use of genetic data but also the population sizes and the cultivated ranges of each genetic unit (Marcer et al. 2016; Milanesi et al. 2018).