Patterns of phylogenetic diversity
Significantly low PD, called "phylogenetic clustering", is likely indicative of habitat filtering. Closely related organisms often share an evolutionarily conservative habitat preference, thus the organisms living together in a particular habitat will tend to be more closely related than expected by chance. Past studies [21, 28, 29, 32] have shown this is especially prevalent in dry habitats, a pattern which is confirmed here. Significantly high PD, called "phylogenetic overdispersion", may be indicative of another ecological process, competitive exclusion. This was only observed here on Taiwan Island, which may be because our study encompasses a very broad area covering several biomes. Future studies focused at a finer scale within a biotic region are more likely to uncover instances of phylogenetic overdispersion.
The discovery of several areas of significantly high RPD in northern, central, and southeastern China (blue areas in fig. 3b) indicates that the flora of those areas contains unusually long branches, i.e. those with relatively few close relatives within China. This could indicate that the flora is relatively old, containing many relictual lineages. In some cases, particularly close to the Chinese borders, it could also indicate the presence of a few lineages of larger clades that mostly occur outside the country, e.g., the extreme southeastern tropical area of the country. The area of significantly low RPD in the middle of the QTP indicates that the flora there contains unusually short branches, potentially indicative of an evolutionary radiation there. All these regions would be worth further studies of processes responsible for the lineages present.
Centers of phylogenetic endemism in northern China
The Tianshan–Altai Mountains, Karakoram Mountains, and Changbai Mountains are boreal mountains located in northern China; these areas and the Kashi area were identified as centers of phylogenetic endemism for the first time in this study. This result adds a crucial complement to the previous non-phylogenetic studies of centers of endemism in China (Additional file 2).
The arid Northwest China biome was influenced by significant climatic oscillations and environmental changes in the Quaternary [43]. There are multiple separate glacial refugia for plants located in the Tianshan–Altai Mountains, such as those for Hippophae [44], Gymnocarpos przewalskii [45], and Reaumuria soonarica [46]. Our results confirmed the above conclusions. The flora of this region was derived from Tethys coastal xerophytes or immigrated from East Asia, and the flora is xeromorphic and ancient [43, 47]. This region’s aridification began in the early Miocene, and drastically increased by desert expansion during the Pleistocene [48, 49]. The lower elevation valleys or mountainous edges of the Tianshan–Altai Mountains likely served as refugia in the Quaternary [43]. During the interglacial periods, refugial lineages may have been able to expand their ranges. Meanwhile, this region’s aridification also stimulated plant diversification [50–53]. These paleobotanical interpretations are supported by our CANAPE results, which showed that relict endemics (rare long branches) and young endemics (rare short branches) are concentrated in the region. In particular, paleo-endemism is concentrated in the Tianshan Mountains, which contains long-branch genera that survived extinction events. The area from Junggar Basin to the Altai Mountains has mainly mixed-endemism. These results indicated co-occurrence in these areas of some relict genera with rare long-branches and young genera with range-restricted short branches. These young genera are concentrated from the Tianshan Mountains to the Altai Mountains and show significantly low PD, especially at the edge regions between mountains and desert. The aridity of this center of endemism might be driving rapid diversification in its core areas of the Gurbantunggut Desert in Junggar Basin [50–53].
The center of phylogenetic endemism discovered in the Changbai Mountains is dominated by paleo-endemism. Our results indicate that these mountains have served as refugia for multiple lineages. Some genera present here, such as Anemarrhena, Mukdenia, and Schisandra, are long isolated branches in the phylogenetic tree, likely because of extinction of close relatives. The preservation of some Tertiary relict plants in the Changbai Mountains refugia was confirmed by previous studies on Asian butternuts (Juglans section Cardiocaryon) [54, 55]. Fossil evidence indicated that the Asian butternuts’ ancestors were widespread in higher latitudes in the northern hemisphere during the early Miocene and then migrated southward because of the late Neogene cooling climates [55]. Extant Asian butternuts are distributed in two large independent refugia, and one of them is in Northeast China with several small-scale refugia scattered across the Changbai Mountains [54].
The linkage of centers of endemism with arid regions seen in this study supports similar findings in other studies. For example, the Sonoran Desert and Chihuahuan Desert were also inferred to have concentrations of paleo-, neo-, super-, and mixed-endemism based on spatial phylogenetic studies of vascular plants in California [21] and Mexico [32]. Likewise, the relatively younger biome of arid areas in northern Chile and the interior of Australia showed mixed- and neo-endemism [28, 29]. Arid and semiarid centers of endemism generally receive less attention and conservation globally because of their lower observed richness at all taxonomic levels, yet the CANAPE approach provides a new view of the unique evolutionary history contributed by xeromorphic lineages.
Centers of phylogenetic endemism in southern China
The Hainan Island center of endemism was relatively distinct in the clustering analysis from the other four centers in southern China (Fig. 5b). The Shigatse and Lhoka Prefecture centers were located in the middle and eastern segments of the Himalaya Mountains, respectively. The two Himalayan centers displayed a close similarity to each other, and clustered at a deeper level with the large center south of the Tropic of Cancer and the Taiwan Island center. The close relationship between these centers indicated that there is close floristic affinity between Taiwan and mainland China, especially southern and southwestern China.
The QTP is the largest, highest, and youngest plateau in the world, and has experienced several uplift events from the early Miocene to the Quaternary [56]. In the QTP, the Himalayan and Hengduan Mountains were previously recognized as areas of high biodiversity and endemism, with many young endemic species [41, 57]. The Shigatse and Lhoka Prefecture centers in Himalaya identified in our CANAPE results are consistent with those recognized by Huang et al. [41] and Zhang et al. [57]. Several previous studies indicated that most endemic genera of the QTP originated in situ or in adjacent regions [58-61], although we detected only one cell significantly dominated by neo-endemism. The significant centers of endemism found here did not include the middle section of the Hengduan Mountains, which has been identified as a center of endemism by previous studies based on traditional taxon-based measures [3, 39–41]. This difference is mainly because we used genera as OTUs in the present study. Several lineages on the QTP have been reported to have undergone extensive radiations in the last million years at the species level [56]; for example, Rhododendron is a species-rich genus of Ericaceae, and many of these species (ca. 159) have restricted distributions in the QTP [62]. Molecular data indicated that the rapid radiation of the Rhododendron subgenus Hymenanthes was driven by hybridization among lineages in the Himalayan region [63–65]. As our study was limited to the genus level, it does not address evolution at the species level. Thus, unlike paleo-endemism, neo-endemism is underrepresented in this study (as discussed in a similar genus-level analysis of the Australian flora by Thornhill et al. [28]). In the future, as more molecular data at species level become available, patterns of neo-endemism can be better estimated.
Southern China is mountainous with a warm and humid subtropical/tropical climate. In our CANAPE analyses, the southern Chinese mountainous areas and Taiwan and Hainan Islands are supported as refugia; these findings were consistent with several other lines of evidence [3, 6, 41].
Conservation concerns
CANAPE can be used to identify areas with significant concentrations of range-restricted long or short branches [26], thus areas harboring relict or new lineages. In our study, nearly all centers of phylogenetic endemism corresponded to mountain ranges. The topographic heterogeneity and relatively stable microclimate of mountains are often cited as being conducive to endemism [1, 6, 7, 66]. Comparing our results with existing nature reserves showed that the majority of the conservation gaps occur in Northwest China (Fig. 6, locations E, F). These conservation gaps of the extensive arid areas include more xeromorphic plant lineages than other areas of China. Additionally, we found that gaps in the Ailao–Jinzhong–Shibalian Mountains should receive conservation attention in the near future (Fig. 6, location B). These conservation gaps are located in the junction of the three mountains and can provide migratory corridors for plants to maintain natural gene flow. Similarly, a conservation gap was found in the Dabie Mountains of central China (Fig. 6, location A), which are considered transition zones for plant exchange [67–69]. Several conservation gaps were also identified in the Shigatse area and Lhoka Prefecture range (Fig. 6, location C and D). Consequently, it is vital to provide more protection to the conservation gaps described above by reducing human disturbance in these areas. Range-restricted taxa require long-term, stable habitats to prosper and persist. Spatial phylogenetic methods allow us to develop approaches for conserving the diversity of rare lineages from an evolutionary standpoint.
Possible limitations due to border effects
Most centers of endemism detected here are near the borders between China and its neighboring countries (Fig. 4). Unlike studies done on biogeographic islands (e.g., Australia [28], Chile [29], and the New Zealand archipelago [70]), our results might be influenced by a political border cutting off a broader floristic region containing lineages with restricted ranges in China that are more broadly distributed across the border. WE and PE are relative only to a given study region, as is indeed true of all biodiversity metrics. Until the data exist to do a world-wide spatial phylogenetic study, this will remain a cause of potential problems with interpretation. Nonetheless, even if some of the relative endemism seen here is only local, it is still significant for biogeographic understanding of the region. Furthermore, it is quite significant for conservation as well, given that countries and other management areas often target organisms that are rare with their boundaries even if present elsewhere.