Aim: In Taiwan island, dispersal through recent land-bridge and oversea after the appearance of proto-Taiwan (<6.5 Ma) is responsible for its biodiversity assembly. While radiations have also been widely reported in mountain systems. So, the tempo-spatial route to the floral hotspot was determined through a meta-analysis of evolution of endemic plants.

Location: Taiwan island and adjacent Asian regions.

Taxon: Seed plants.

Methods: Published dated phylogenies were compiled, and the stem ages were considered as origin times. Ancestral range shift pattern and/or distribution of sister lineage/species were used to determine the speciation mode, in situ origin or dispersal. The distributions of sister taxa were assigned to ten different biogeographical regions. Histograms quantifying the number of different origin mode events within 0.5 Ma bins were constructed. Possibility of long-distance dispersal (LDD) was evaluated.

Results: The 125 sampled endemic plants originated between the late Eocene and late Pleistocene, with 111 species (88.8%) originating after the emergence of proto-Taiwan and 14 species (11.2%) before. Spatially, in situ speciation contributed more to the formation of endemism than dispersal (56.8% vs 43.2%). The mean ages of in situ origin species (1.83 Ma) are significantly younger than that of dispersal (6.34 Ma). The main regions from which dispersal occurred were south-central (n=29) and southeast China (n=28) (referred to as south China), followed by Japan (n=14). High ratio (60.8%) of LDD ability is found.

Main conclusions: Spatially, recent land-bridge and oversea dispersal from the adjacent flora of Asia is supported and southwest and southeast China are the most important sources. In situ speciation that may be correlated with mountain uplift, monsoon intensification and Pleistocene climatic fluctuations exert greater contributions. Temporally, floral endemism is predominantly origin after the emergence of proto-Taiwan. Additional investigations with more sophisticated sampling and genetic data are needed in the future.