How did climate drive the historical distribution of relict plants: a case study on Metasequoia using deep-time ecological niche modelling Authors: Junmei Chen, Ai Song, Li Wang, Shufeng Li, Tao Su Published: 2026 ABSTRACT Relict plants are those that once had a wide distribution but are now restricted to small areas. It is important to study the process of their distribution in the geological past to better understand the climatic factors that shape their current distribution. Metasequoia is one of the best-known genera of relict plants in Asia, with rich fossil records dating back to the Late Cretaceous. In this study, we used Deep-time Ecological Niche Modelling (DENM) and 755 Cenozoic megafossil records of Metasequoia to understand the pattern of its historical distribution and driving factors. The results showed that Metasequoia was widely distributed across mid-high-latitude regions during the Paleogene. Its distribution range contracted toward lower latitudes during the Neogene due to global aridification and cooling. Our results indicate that mean annual precipitation was the key factor shaping the distribution of Metasequoia during the Paleogene and Neogene. Later during the Pleistocene, a dramatic contraction of its distribution occurred due to glaciations, which also shaped the modern distributions of many relict plants in China. Our study supports that climate change is the main factor driving the distribution of relict plants and therefore provides crucial evidence for understanding their biogeographic histories and future conservation under climate change. Introduction The term “relict plant” refers to species that were once widespread in the geological past but are now endemic to limited geographic regions (Habel et al., 2010). East Asia has long been regarded as refuge for many relict plants (Qiu et al., 2017); however, the mechanisms driving the modern distributions of relict plants are largely unresolved. Previous studies have suggested that climate plays an important role in their modern distributions (Huang et al., 2015; Tang et al., 2018), but there is still limited understanding for the histories of their distribution. Metasequoia glyptostroboides Hu et W. C. Cheng is one of the most well-known extant species among relict plants and it is the sole species in Metasequoia. The modern distribution of Metasequoia is naturally limited to western Hubei, northern Hunan, and Chongqing in China with around approximately 60,000 hectares (Figs. 1 and 2) (Zheng & Fu, 1978; Leng et al., 2007; Huang et al., 2020). Present and fossil specimens of Metasequoia. (A) The cone of Metasequoia glyptostroboides. (B) The branch of M. glyptostroboides. (C) The foliaged shoot of M. glyptostroboides. The photos of living specimens are of cultivated trees in Chengdu. (D) The Middle Miocene fossil of Metasequoia sp. (cf. Metasequoia glyptostroboides) from Sanzhangtian in Yunnan, China. (A, C, and D) Scale bars: 1 cm. (B) Scale bar: 5 cm. The global Cenozoic fossil sites of Metasequoia and their current locations. Current locations obtained from GBIF (https://doi.org/10.15468/dl.j6p8sx) and manually verified using google earth. The Cenozoic fossil sites are based on research by LePage, Yang & Matsumoto (2005) and the Paleobiology Database (https://paleobiodb.org), as well as literature not included in these datasets (Qi, Yang & Su, 1993; Greenwood et al., 2005; Williams et al., 2008; Moiseeva, Herman & Spicer, 2009; Smith, Basinger & Greenwood, 2009; Parrish et al., 2010; Pigg & DeVore, 2010; Herrera et al., 2011; Sunderlin et al., 2011; Paik et al., 2012; Stockey, Hoffman & Rothwell, 2013; Greenwood, Pigg & DeVore, 2016; Ding et al., 2018; Li, Jin & Manchester, 2018; Lowe et al., 2018; Wang et al., 2019; Pavlyutkin et al., 2020; Nishino et al., 2023; Pigg, Manchester & Devore, 2023; Fukushima, Suzuki & Momohara, 2024; Yamakawa et al., 2024; Liu et al., 2025). The base map was created using Natural Earth (2024). Metasequoia has quite an abundant fossil records and is widely distributed in the Northern Hemisphere. The earliest occurrence of Metasequoia has been dated to the Early Cenomanian (Late Cretaceous) in British Columbia, Canada (LePage, Yang & Matsumoto, 2005). The first fossil report of Metasequoia was documented in the Chuckanut Formation near Birch Bay, Washington, USA, which was once classified as Taxodium occidentale (Newberry, 1863). Subsequently, Miki (1941) established a new genus, Metasequoia, based on fossil material from central Hondo, Japan. According to the International Fossil Plant Names Index (https://ifpni.org/index.htm), 35 fossil species of Metasequoia have been reported. Following a thorough morphological comparison, however, Liu, Li & Wang (1999) assigned the majority of these fossil species to Metasequoia occidentalis according to their close morphological similarity. In addition, Stockey, Rothwell & Falder (2001) designated Metasequoia milleri as a separate fossil species with distinct anatomical features. Previous fossil records throughout the Cenozoic have indicated that Metasequoia ranged from the mid-latitudes to the Arctic. It has been identified as one of the most prevalent components of Northern Hemisphere flora since the Early Paleocene (LePage, Yang & Matsumoto, 2005). However, during the Paleocene-Eocene Thermal Maximum (PETM), the distribution of Metasequoia expanded and these species became a major component of Arctic flora (Leng & Yang, 2009). During the Pliocene, Metasequoia disappeared from most parts of Central Eurasia and North America (Liu, Arens & Li, 2007). During the Quaternary, the spatial distribution of Metasequoia contracted dramatically and it finally naturally survives only in several humid valleys of central China (Chu & Cooper, 1950; Leng et al., 2007). Despite these fossil records document the historical distribution of Metasequoia, the mechanisms underlying patterns and driving factors of dispersal still have not been fully investigated. Ecological Niche Modeling (ENM) is a reliable approach for interpreting biogeographic patterns that is largely applied in the biological sciences (Meseguer et al., 2015; Olivera, Minghetti & Montemayor, 2021; Campos et al., 2023; Thuiller, 2024; Wen et al., 2024). This method is used to extract environmental variables for native occurrence sites of a studied species and map their potential distribution regions in the past and future (Thuiller, 2024). Deep-time Ecological Niche Modeling (DENM) is a methodological framework that applies ENM to paleoecological studies. This methodology has been shown to be highly reliable for bioclimatic factor analysis, and research incorporating fossil records and biogeographic analysis can shed new light on the factors driving geohistorical changes in distribution (Meseguer et al., 2015; Myers, Stigall & Lieberman, 2015; Chen, 2021; Li et al., 2022; Gao et al., 2023). In this study, we used ENM and the current native occurrence of Metasequoia to identify the most effective bioclimate factors shaping its distribution. We applied these factors to global fossil records, in conjunction with DENM, to reconstruct the historical distribution of Metasequoia throughout the Cenozoic. Meanwhile, we analyzed the climatic factors that shaped its historical distribution during different geological periods. This study can provide substantial evidence on the responses of relict plants to past global climate change, as well as insights into plant conservation under climate change in the future. Materials and Methods Modern distribution and Cenozoic fossil records of Metasequoia Metasequoia glyptostroboides presence occurrences were obtained from the Global Biodiversity Information Facility (https://doi.org/10.15468/dl.j6p8sx). The coordinates file was transferred to a KML file using the writeOGR function in R by the “rgdal” package (version 1.6.7) (Bivand, Keitt & Rowlingson, 2023). Only the regions of western Hubei, northern Hunan, and Chongqing were considered to be its modern natural distribution (Leng et al., 2007; Huang et al., 2020), records in these areas were used in the analysis. We adopted the age and location of each megafossil record from LePage, Yang & Matsumoto (2005) and the Paleobiology Database (https://paleobiodb.org), and fossil records were supplemented by a comprehensive review of the published literature (Data S1) (Qi, Yang & Su, 1993; Greenwood et al., 2005; Williams et al., 2008; Moiseeva, Herman & Spicer, 2009; Smith, Basinger & Greenwood, 2009; Parrish et al., 2010; Pigg & DeVore, 2010; Herrera et al., 2011; Sunderlin et al., 2011; Paik et al., 2012; Stockey, Hoffman & Rothwell, 2013; Greenwood, Pigg & DeVore, 2016; Ding et al., 2018; Li, Jin & Manchester, 2018; Lowe et al., 2018; Wang et al., 2019; Pavlyutkin et al., 2020; Nishino et al., 2023; Pigg, Manchester & Devore, 2023; Fukushima, Suzuki & Momohara, 2024; Yamakawa et al., 2024; Liu et al., 2025). Previously, the ages of Metasequoia fossil records were determined by evidence mainly from radiometric dating, paleomagnetism, mammal fossils and palynological comparison. If the age of a fossil record was assigned to more than one stage (e.g., Late Eocene to Early Oligocene), it was considered as occurred in these stages for simulation. After this treatment, seven hundred and fifty-five Cenozoic fossil records were compiled. We used the R package “chronosphere” (version 0.6.1) and Gplates (https://gws.gplates.org) (Kocsis & Raja, 2013; Müller et al., 2018) to reconstruct paleocoordinates for each fossil record. Climate data The climate data for both modern and fossil records were derived from the HadCM3B model version 2 (https://www.paleo.bristol.ac.uk/resources/simulations/). The production of this dataset was overseen by the UK Hadley Centre/Meteorological Office, with subsequent development for the specific, regular use of the Bristol Research Initiative for the Dynamic Global Environment group (http://www.bristol.ac.uk/geography/research/bridge). The selected model is an atmosphere-ocean General Circulation Models (GCMs) that is based on the HadCM3 atmosphere model and coupled with an atmosphere-ocean GCM (Valdes et al., 2017; Valdes, Scotese & Lunt, 2020). The model resolution is 96 × 73 grid points, equivalent to 3.75° longitude × 2.5° latitude (Valdes et al., 2017). The dataset extracted from the original model contains 14 bioclimate variables in a group for each epoch as asc files for standalone use with latitude and longitude. Maximum entropy modeling The most significant bioclimate variables that shape the modern distribution of M. glyptostroboides were determined using Maximum entropy (MaxEnt) (version 3.4) (Elith et al., 2011; Phillips, Dudík & Schapire, 2024). The simulation was replicated 50 times and exported as logistic-created response curves. The jackknife method was employed to ascertain the importance of the variables. The remaining settings were configured to utilize the default attributes. Model performance was evaluated using the area under the receiver-operator curve (AUC) with values ranging from 0 to 1. It is generally accepted that higher AUC values indicate a more reliable result (Elith et al., 2006). The contribution of each climate variable was evaluated and the correlation among disparate variables were calculated. In the following steps we chose eight bioclimate variables that have the highest contributions based on the MaxEnt simulation and the Pearson correlation analysis used function layerStats within package “spatialEco” (version 2.0.2) in R (Fig. S1) (Evans & Murphy, 2023). Then, the eight selected bioclimate variables were simulated with MaxEnt for another round to determine their contribution. Deep-time ecological niche modelling The extant and geological distribution of Metasequoia, as well as its climate niche, have undergone substantial changes. In this study, the Deep-time ecological niche modelling (DENM) method was employed to analyze ecological niches over geohistorical time scales. The conceptual foundation of DENM was predicated on integrating the fossil record with ENM into biogeographical studies (Meseguer et al., 2015). The bioclimate factors chosen from the MaxEnt contribution and correlation analysis were then input into the DENM model to simulate the past potential distribution regions. Firstly, the Extract function in the R package “raster” (version 3.6.30) was used to extract eight bioclimate variables from fossil record paleo-coordinates for each epoch (Hijmans et al., 2024). Thereafter, the overlap geographical regions corresponding to the eight bioclimatic ranges of the same epoch were extracted using the intersection layer extraction, and a distribution mask was constructed using the Mask function in R with the “raster” package (version 3.6.30) (Hijmans et al., 2024). Subsequently, the intersection and covariance values for the eight bioclimatic variables within the designated geographical overlap region were calculated from bioclimatic data. Intersection regions were the most suitable climatic condition for Metasequoia during that epoch. In this research, Mahalanobis distance was calculated using an R script due to the uneven distribution and incompleteness of the fossil record. The Mahalanobis coefficient, a statistical measure of multivariate data, ranges from 0 to 1. A value of 0 indicated an area of potential growth for Metasequoia, while a value of 1 signified an area deemed unsuitable for growth. The Mahalanobis binary was calculated using the Mahalanobis function in the R programming language with the “stats” package (version 4.3.3) (R Core Team, 2024). Results Bioclimate factors for the modern distribution of Metasequoia glyptostroboides A Maxent simulation was conducted to estimate the potential distribution area of Metasequoia. The model had a mean AUC of 0.978, indicating high reliability. The analysis indicated that the most significant bioclimate variables influencing the Metasequoia distribution were mean annual precipitation (MAP) (34.470%) and wettest month precipitation (WETMON) (23.115%). The variables that exerted less influence on Metasequoia distribution were as follows: driest season precipitation (DRYSEA) (13.548%), wettest season-driest season precipitation difference (WETDRYSEA) (13.080%), warmest season-coldest season temperature difference (WSMCSM) (6.813%), warmest month mean surface air temperature (WMM) (6.546%), driest month precipitation (DRYMON) (2.224%), and coldest month mean surface air temperature (CMM) (0.204%) (Fig. 3). The contribution of different bioclimate variables. Mean annual precipitation (MAP) (34.470%), Wettest month precipitation (WETMON) (23.115%), Driest season precipitation (DRYSEA) (13.548%), Wettest-driest season precipitation difference (WETDRYSEA) (13.080%), Warmest-coldest season temperature difference (WSMCSM) (6.813%), Warmest month mean (WMM) (6.546%), Driest month (DRYMON) (2.224%), Coldest month mean (CMM) (0.204%). Reconstruction of bioclimate variables throughout the Cenozoic Eight groups of bioclimate data were extracted from the locations of 755 fossil records and present occurrences (Fig. 4). The data demonstrated that the ranges of bioclimatic variables related to temperature remained relatively stable until the Pleistocene. Conversely, there was a significant decline to a limited bioclimatic range until the Pleistocene (Fig. 4). This phenomenon partially corresponded to the distribution change of Metasequoia (Fig. 5). The variability of different precipitation-related variables was significantly reduced during the Oligocene and increased during the Miocene; whereas variables related to temperature were not so prominent during that time span. A substantial percentage that was greater than 50% of the extreme values for biological variables of Metasequoia emerged during the Miocene, indicating that the Miocene was one of the most significant periods shaping the distribution of Metasequoia (Data S2). A dramatic change in bioclimate variables from the Pleistocene was evident in the extracted data, indicating that the climate exerted significant influence on the distribution area during that period. The range of variations in bioclimate variables associated with the growth of Metasequoia has been observed to change during different epochs. Mean annual precipitation (MAP), wettest month precipitation (WETMON), driest season precipitation (DRYSEA), wettest season driest season precipitation difference (WETDRYSEA), driest month (DRYMON), warmest season coldest season temperature difference (WSMCSM), warmest month mean (WMM), and coldest month mean (CMM). The latitude range and distribution location changes in different epochs during the Cenozoic. The fossil record uses the paleocoordinates generated by GPlates (Müller et al., 2018). The pink points represent the median values of each epoch. Reconstruction of distribution regions throughout the Cenozoic The reconstruction results showed that Metasequoia was widespread throughout the Northern Hemisphere during the Paleogene (Fig. 6A) and reached its maximum geographic range during the Paleocene-Eocene Thermal Maximum (PETM) (Figs. 6A and 6B). After this, there were two significant contractions in geographic distribution: one during the Eocene-Oligocene transition (EOT), particularly evident around the Arctic Ocean and Greenland (Figs. 6B and 6C); the other during the Pliocene-Pleistocene transition period, affecting most of the Northern Hemisphere (Figs. 6F and 6G). During the Neogene, the northern boundary of its potential distribution range shifted southward. In comparison, the southern boundary of its potential distribution range shifted southward until the Miocene, followed by a slight northward shift in the subsequent period (Fig. 5). The northernmost extent of the historical distribution region of Metasequoia extended from the Arctic Ocean during the Pliocene. Metasequoia migrated to the lowest latitudes during the Pleistocene, a phenomenon that closely resembled its contemporary native distribution (Figs. 5 and 6G). The distribution range of Metasequoia during various geological periods in the Cenozoic. (A) Paleocene, (B) Eocene, (C) Oligocene, (D) Early-Middle Miocene, (E) Late Miocene, (F) Pliocene, (G) Pleistocene, (H) Present. The regions exhibiting the highest levels of suitability are delineated in red, while those demonstrating the least suitability are indicated in dark blue. The definition of climatically suitable areas was based on the Mahalanobis distance, which was set at a maximum of 0.2. The extant bioclimate data are devoid of any information pertaining to Antarctica. The Antarctica map is Pleistocene in origin, as the geography is almost identical and not conducive to the growth of Metasequoia. Base map is modified from HadCM3 (Valdes et al., 2017). Moreover, the quantitative analysis of the potential distribution of Metasequoia across different epochs further revealed two significant declines (Fig. 7). These declines occurred during the Eocene and Oligocene, as well as during the Pliocene and Pleistocene. However, there was a marked decrease in the potential distribution in the Southern Hemisphere, where no Metasequoia fossil record has been documented (Figs. 6B–6D). It is noteworthy that the potential distribution during the Pleistocene was highly similar to the present natural occurrence area. Potential distribution area changes of Metasequoia in different epochs. Discussion Geohistorical distribution of Metasequoia This study combined fossil records with DENM to reconstruct the historical distribution of Metasequoia. Our results demonstrated that the potential distribution regions throughout the Cenozoic were generally divided into three stages (Fig. 6). The initial stage was from the Paleocene to the Oligocene, during which Metasequoia was the dominant species in nearly the entire Northern Hemisphere north of 25°N. The second stage, spanning from the Miocene to the Pliocene, was characterized by a progressive contraction of the potential distribution region. The third stage was marked by a significant spatial contraction since the Pleistocene, a period in which their distribution was limited to the Japanese archipelago and southeastern China. The current natural distribution of Metasequoia is confined to mountainous valleys in western Hubei, northern Hunan, and Chongqing of Central China. Since the Paleocene, the Metasequoia distribution has expanded extensively in the Northern Hemisphere, reaching as far north as 82°N (Data S1) and extending to the boundaries of the Arctic Ocean (Fig. 6A) (LePage, Yang & Matsumoto, 2005). During the PETM, the Arctic experienced a significant global increase in temperature and humidity (Westerhold et al., 2020; Rush et al., 2021; Tierney et al., 2022). The potential distribution regions of Metasequoia reached a peak. Meanwhile, Eocene fossil records have not shown a significant change in distribution compared to the Paleocene (Figs. 6A and 6B). The distribution of Metasequoia clearly contracted in high latitudes between the Eocene and Oligocene (Figs. 6B and 6C). The closure of the Arctic Ocean decelerated ocean currents and altered the precipitation patterns of the Pacific and Atlantic Rims, which, combined with significant global temperature decreases during the EOT, intensified seasonality at high latitudes (Ivany, Patterson & Lohmann, 2000; Eldrett et al., 2009; Westerhold et al., 2020). The expansion of consistent ice sheets at high latitudes, especially around Greenland, was mainly driven by hydrographic conditions that also caused a reduction in fluvial inputs (Kinnard et al., 2008; Tripati & Darby, 2018; Inglis et al., 2020). During the Oligocene, continental glaciers and the limited presence of sediments led to an absence of Metasequoia records around North Greenland (LePage, Yang & Matsumoto, 2005; Fujita, 2008). Fossil records from North America have indicated that Metasequoia moved westward; its occurrences in East Asia are relatively stable. The change in Metasequoia distribution in central inland region of Eurasia followed the closure of the Turgai Strait (LePage, Yang & Matsumoto, 2005; Averyanova et al., 2024). From the Oligocene to the Early-Middle Miocene, the potential distribution regions of Metasequoia migrated southward and were accompanied by a remarkable decline (Figs. 6C and 6D). A significant decline in Metasequoia fossil records during the Late Miocene indicated that this was a pivotal transition for its distribution decline (Figs. 6D and 6E). During the Middle Miocene, the closure of the Tethys Ocean and a consequent decrease in sea level caused by the permanent glaciation of Antarctica may have induced regional drought in the inland region of Eurasia (Bialik et al., 2019; Xu et al., 2023). Consequently, there was a pronounced decline in the spatial distribution of Metasequoia. There was a slight change in potential distribution regions between the Late Miocene and the Pliocene (Garcia-Castellanos et al., 2009; Ao et al., 2021), however the Metasequoia fossil records had a significant decline (Figs. 6E and 6F). Significant declines in both fossil records and potential distribution regions were observed between the Pliocene and the Pleistocene (Figs. 6F and 6G). In North America, the climate transitioned from being relatively warm and humid to a drier, glacial state during the Pliocene-Pleistocene transition, which led to the disappearance of Metasequoia there (Ehlers, Gibbard & Hughes, 2011; Lang et al., 2014). Glaciers distributed closely to the potential distribution area of Metasequoia in Japanese archipelago during the Last Glacial Maximum, witch possible lead to the disappearance of this genus after this event (Ehlers, Gibbard & Hughes, 2011) (Fig. 6G). Today, Metasequoia is only found in restricted areas of central China (Figs. 6G and 6H). Climatic factors driving the historical distribution of Metasequoia According to our analysis using Maxent, precipitation was identified as a major factor in shaping the historical distribution of Metasequoia; specifically, mean annual precipitation (MAP) and wettest month precipitation (WETMON) were the most important precipitation variables (Fig. 3). The historical distribution pattern of Metasequoia matched well with the change in regional precipitation. For example, the closure of the Neo-Tethys led to the aridification from southeastern Europe to East Asia during the Miocene (Ivanov et al., 2011; Sun et al., 2021; Xu et al., 2023), which resulted in massive vegetation turnover. Metasequoia was once flourishing there and subsequently disappeared (Figs. 6C and 6D). Moreover, in North America, the closure of the Arctic Ocean caused a decrease in precipitation around the Atlantic Rim and an increase in precipitation around the Pacific Rim (Hutchinson et al., 2019), which led to the Metasequoia distribution from central to western North America (Figs. 6C and 6D). Later, the Middle Miocene drought in North America caused the expansion of C4 vegetation (Loughney, Harkness & Badgley, 2023) and accelerated the regional disappearance of Metasequoia. Following MAP and WETMON, driest season precipitation (DRYSEA) and wettest season-driest season precipitation difference (WETDRYSEA) were subsidiary effects that indicated seasonality (Fig. 3). Since the Late Miocene in East Asia, the continuous uplift of the Tibetan Plateau has been a primary factor in the strengthening of the monsoon system, leading to alterations in humid circulation patterns (Tada, Zheng & Clift, 2016). These changes resulted in drier winters and earlier springs, which in turn limited the population establishment of young sprouts (Wang et al., 2019; Fan et al., 2020). Through the geohistorical distribution in Cenozoic indicated that fossils were only found in oceanic climates with sufficient moisture or monsoonal and Mediterranean climates with adequate rainfall during the spring months, which was consistent with the current distribution of M. glyptostroboides (Gushulak, West & Greenwood, 2016; Fan et al., 2020; Matsuzaki, Ikeda & Tada, 2022; Fukushima, Suzuki & Momohara, 2024; West et al., 2024). In addition, we found that the distribution range of Metasequoia was not significantly impacted by temperature until the occurrence of the Pleistocene glaciers (Fig. 6). For example, North America experienced a decrease in temperatures during the Eocene-Oligocene Transition (Zanazzi et al., 2007) but the distribution of Metasequoia remained relatively stable along those latitudes (Fig. 5). The Glacier covered North America during the Pleistocene (Clark et al., 2009) probably lead to the disappearance of Metasequoia in this region (Fig. 6G). Glaciers occurring during the Last Glacial Maximum (Sawagaki & Aoki, 2011) might have been the reason for the disappearance of Metasequoia from the Japanese archipelago (Figs. 6G and 6H). Additionally, the massive orogenic events and volcanic activity that started from the Late Miocene to Early Pleistocene blocked the migration of Metasequoia to southern Japan (Momohara, 2005). Meanwhile, Metasequoia survives in central China with complex topography that provides refuges under a humid environment (Tang et al., 2018). Conclusion This study indicated that the historical distribution of Metasequoia throughout the Cenozoic can be summarized into three stages: (1) a relatively stable distribution during the Paleogene; (2) a significant southward shift during the Neogene; and (3) a dramatic reduction during the Pleistocene, forming its modern distribution. By integrating DENM and numerous fossil records, we propose that from the Paleogene to the Neogene, the key bioclimatic variable influencing the distribution of Metasequoia was precipitation, as its distribution gradually contracted with regional decreases in precipitation. However, Quaternary climate changes, particularly decreases in temperature that were accompanied by glaciations, further contributed to the sharp decline in the distribution range of Metasequoia. Finally, Metasequoia survives today only in restricted areas of central China with complex topography and humid conditions. Generally, this study not only provides important evidence for understanding the historical distributions of Metasequoia and its climatic driving factors, but also implies that future climate changes, and particularly drought, may have significant impacts on the survival of relict plants. Supplemental Information Metasequoia Cenozoic distributions. Historical distribution of Metasequoia fossil records from different epochs. 8 extract bioclimate variables from different stages in Cenozoic. 8 groups of bilclimate varibles extracted from the Metasequoia historical occorence in different epoches used in Figure 4. Pearson correlation. Result of 14 modern bioclimate variables. Mean annual precipitation (MAP). Figure 4 with detail and p-values Wettest month precipitation (WETMON). Figure 4 with detail and p-values Driest season precipitation (DRYSEA). Figure 4 with detail and p-values Wettest season-driest season precipitation difference (WETDRYSEA). Figure 4 with detail and p-values Warmest season-coldest season temperature difference (WSMCSM) (6.8%). Figure 4 with detail and p-values. Warmest month mean surface air temperature (WMM). Figure 4 with detail and p-values. Driest month precipitation (DRYMON). Figure 4 with detail and p-values. Coldest month mean surface air temperature (CMM). Figure 4 with detail and p-values. Maxent simulation result with full 14 selected bioclimate variables. Maxent simulation result with 8 selected bioclimate variables.