Species delimitations by morphological and by genetic markers are not always congruent. Magnolia kobus consists of two morphologically different varieties, kobus and borealis. The latter variety is characterized by larger leaves than the former. For the conservation of M. kobus genetic resources in natural forests, the relationships between morphological and genetic variation should be clarified. We investigated variations in nuclear microsatellites, chloroplast DNA (cpDNA) sequences and leaf morphological traits in 23 populations of M. kobus over the range of species. Two genetically divergent lineages, northern and southern were detected and their geographical boundary was estimated to be at 39°N. The northern lineage consisted of two genetic clusters and a single cpDNA haplotype, while the southern one had multiple genetic clusters and cpDNA haplotypes. The northern lineage showed significantly lower genetic diversity than the southern. Approximate Bayesian computation indicated that the northern and southern lineages had experienced, respectively, population expansion and long-term stable population size. The divergence time between the two lineages was estimated to be 565,000 years ago and no signature of migration between the two lineages after divergence was detected. Ecological niche modeling showed that the potential distribution area in northern Japan at the last glacial maximum was very small. It is thus considered that the two lineages have experienced different population histories over several glacial-inter-glacial cycles. Individuals of populations in the central to northern part of Honshu on the Sea of Japan side and in Hokkaido had large leaf width and area. These leaf characteristics corresponded with those of variety borealis. However, the delimitation of the northern and southern lineages detected by genetic markers (39°N) was not congruent with that detected by leaf morphologies (36°N). It is therefore suggested that variety borealis is not supported genetically and the northern and southern lineages should be considered separately when identifying conservation units based not on morphology but on genetic markers.