Geogenomics seeks to understand geological processes linked to lineage divergence. However, the mechanisms that conserve ancient signals despite gene flow are still unclear. In the southern beech, the deep lineage divergence produced by vicariant events is associated with ancient marine transgressions. We hereby evaluate the hypothesis that this divergence is maintained by diversifying selection. The lineage divergence using AMOVA, principal coordinate analysis, assignment tests, and multiple matrix regression analyses was assessed using chloroplast DNA and neutral and outlier SNPs. Several environmental variables were used to characterize potential within-species niche structuring and genotype-environment associations. Two deep-rooted latitudinally structured lineages resulted from cpDNA, the northern cluster being more genetically diverse than the southern one. Of the total of 2,943 SNPs, 33 were identified as outliers and produced two genetic clusters. Neutral SNPs yielded no structure by AMOVA, whereas higher (>75%) F_st values were obtained for cpDNA and outlier SNPs. Precipitation variables were mostly associated with population clusters and suggested two climatic niches, consisting of cold and dry in the south and more variable precipitation, temperature, and soil conditions in the north. Associations of genetic distance with environment and geography suggested IBD and IBE effects. Ancient lineage divergence in N. dombeyi, originally driven by vicariance, has been maintained by diversifying selection under distinct environmental conditions that also define distinct within-species niches. Deeply rooted phylogeographic breaks can be conserved in continuously distributed species in the absence of current geographic barriers. Yet physical gradients exert differential selective pressures, which are maintained in the face of potential gene flow. As a result, selection can lead to geographically localized and differentially adapted groups of populations that can be detected by a combination of traditional phylogeographic and novel genomic methods.