1. Introduction

A central goal of biogeography is to explain the heterogeneous patterns of biodiversity across space and time (Cox et al., 1976; Sanmartín, 2012). These patterns are shaped by biogeographical processes, such as dispersal and extinction, and by niche evolution (Zhang et al., 2021a, 2021b; Li et al., 2025). Global species diversity is highly heterogeneous, and even within the same clade, significant disparities can be observed. For example, within Juglandaceae, the subfamily Juglandoideae comprises approximately 80% of the family’s species, most of which inhabit temperate deciduous forests in the Northern Hemisphere. In contrast, Engelhardioideae is predominantly distributed in subtropical to tropical forests (Zhang et al., 2021b). Similarly, major niche divergence has been documented among different clades of Juniperus (Liu et al., 2024). These patterns arise from complex interactions between geological history, climatic shifts, dispersal capacity and niche differentiation (Coelho et al., 2023; Sekar et al., 2023; Song et al., 2024). Historical climate changes have forced plant species to migrate, adapt their ecological niches, or face extinction (Wolfe, 1975; Tiffney, 1985a; Wen, 1999; Manos and Meireles, 2015). The mechanisms drivingthese disparities and their impact on biodiversity remain subjects of ongoing investigation by systematists and paleobotanists (Tiffney, 1985a, 1985b; Wen, 1999; Wang et al., 2010; Zhang et al., 2018; Qian et al., 2019).

The advancement of sequencing technologies has enabled researchers to make remarkable progress in reconstructing the biogeographical history of lineages by utilizing genetic information from extant species. Phylogenomic frameworks based on extant taxa have successfully resolved divergence times, historical range evolution, and broad diversification patterns. These approaches have proven particularly valuable for several groups, such as many herbaceous plants (Ramírez et al., 2007; Givnish et al., 2015, 2018; Serna–Sánchez et al., 2021; Ling et al., 2024; Wang et al., 2024). However, the reliability of extant–only approaches declines markedly for ancient lineages that have experienced extensive extinction, range contraction, or repeated intercontinental dispersal.

Fossil records, as direct evidence of past life, preserve temporal, spatial, and ecological data essential for biogeographic reconstructions, playing an irreplaceable role in elucidating floristic dynamics (Manchester, 1999; Tiffney and Manchester, 2001; Manchester et al., 2009). First, fossils provide critical temporal calibration points, substantially improving divergence time estimates when integrated into phylogenetic frameworks of extant clades (Chen et al., 2017; Zhang et al., 2021a). For instance, fossil–integrated phylogenetic analyses have greatly refined divergence–time estimates in diverse animal groups, including mammals and insects (López–Antoñanzas et al., 2024; Wutke et al., 2024; Wang et al., 2025). Second, by incorporating fossils as independent branches or node attributes based on their systematic positions and geographic occurrences, they enhance the dimensionality of historical biogeographic analyses (Li et al., 2025). In woody plant lineages, numerous studies have shown that integrating macrofossils or leaf physiognomic data can fundamentally reshape inferences of lineage diversification, climatic niche evolution, and ancient floristic exchanges, as demonstrated in Cycadales, Juniperus, and Fraxinus (Coiro et al., 2023; Li et al., 2024; Liu et al., 2024). More broadly, constructing a comprehensive model of species diversity pattern evolution requires an interdisciplinary framework. The synergistic integration of molecular phylogenetics with fossil evidence (total–evidence dating), combined with paleoclimate reconstructions (Williams and Jackson, 2007), overcomes the spatiotemporal resolution constraints of single–data approaches. For ancient lineages, fossil–integrated studies have become essential for uncovering the mechanistic drivers of their diversity (Li et al., 2024).

The Ulmaceae family is a lineage with abundant and well–preserved fossil records, consisting of seven genera and 58 species, is widely distributed across the globe and represents a key component of the northern temperate flora, with substantial ecological importance (Kozlowski et al., 2018; Fragnière et al., 2021; Zhang et al., 2021a). East Asia represents the primary center of species diversity in Ulmaceae, followed by North America, while Europe displays the lowest species richness (Fragnière et al., 2021). Based on integrated morphological and molecular phylogenetic evidence, Ulmaceae is divided into two major taxonomic and biogeographical clades: a temperate clade and a tropical clade. The temperate clade includes Ulmus, Zelkova, Hemiptelea, and Planera, which are predominantly deciduous species distributed widely across mid–to–high latitude regions of the Northern Hemisphere. In contrast, the tropical clade consists Ampelocera, Phyllostylon, and Holoptelea, which are largely evergreen and extend into tropical regions (Fu, 1980; Sherman–Broyles et al., 1997; Fu et al., 2004; Neubig et al., 2012; Fragnière et al., 2021; Zhang et al., 2021a). Notably, the temperate clade exhibits higher species richness than the tropical clade, and the two clades diverge significantly in their climatic niche preferences. Zhang et al. (2021b) combined phylogenomic and fossil evidence to investigate the pronounced differences in species diversity between the two clades, and emphasized the roles of historical biogeography, and climatic change in shaping the modern distribution of Ulmaceae. However, the impact of niche evolution on the diversity patterns of Ulmaceae species and the causes underlying the asymmetric diversification remain unclear.

In this study, we investigate the factors contributing to the pronounced disparity in species diversity distribution between the temperate and tropical clades of Ulmaceae. Using a heuristic approach, we integrated fossil taxa into the phylogeny based on taxonomic classification and estimated ages, even when their precise phylogenetic positions were uncertain. We constructed a robust time–calibrated species–level phylogeny encompassing 86.21% of extant species and 61 fossil species. By combining molecular phylogenetics, fossil evidence, and paleoclimate reconstructions, this study aims to elucidate how geographic isolation, historical dispersal, and ecological niche evolution collectively shaped the species distribution patterns of the two major clades within Ulmaceae.

2. Materials and methods 2.1. Sample collection and data compilation

According to the recently published papers and floras (Fu, 1980; Fragnière et al., 2021; Zhang et al., 2021a), we adopted Ulmaceae containing 58 extant species (Table S1). In this study, we collected 50 species of Ulmaceae, representing all seven extant genera and approximately 86.21% of the extant species. Twenty–six species were collected from the field and herbarium specimens (Museum of Beijing Forestry University). Chloroplast genome sequences of the other 24 species were downloaded from GenBank. Additionally, five species from Cannabaceae were selected as the outgroups (Tables S2 and S3) (Group et al., 2016).

The distribution data of Ulmaceae species were obtained from online databases, including the Global Biodiversity Information Facility (GBIF), National Plant Specimen Resource Center (NPSRC), NSII (http://www.nsii.org.cn), and peer reviewed papers and published floras (Fu, 1980; Fragnière et al., 2021; Zhang et al., 2021a). We rigorously screened these data to remove records with cultivated specimens, duplicate records and some erroneous records (i.e., occurrences in the oceans and deserts). Only one occurrence record within 10 km was reserved for reducing the effects of spatial autocorrelation and the consequent overfitting. Our final dataset contained a total of 36,691 occurrence records from all 58 species, and we set a minimum threshold of four records per species (Table S4).

Nineteen bioclimatic variables of near–current climate (representative of 1970–2000) were downloaded from the WorldClim 2.1 (https://www.worldclim.org/) at a 2.5 arc–min spatial resolution (Table S5) (Fick and Hijmans, 2017). To reduce collinearity, we tested the correlation based on the distribution points of extant species and present climate variables from WorldClim 2.1. The relative contribution and permutation importance of all 19 bioclimatic variables were first evaluated using a Maximum Entropy (MaxEnt) model based on extant species occurrences, in order to quantify the explanatory power of each variable for present–day species distributions (Table S6). Pearson’s correlation tests were then applied to assess collinearity among variables, and highly correlated variables (|r| > 0.8) were removed following variable ranking by contribution. The final set of relatively independent and informative variables retained for niche–related analyses comprised Bio1, Bio3, Bio9, Bio12, Bio17, and Bio19 (Tables S6 and S7).

We compiled a comprehensive dataset of Ulmaceae fossils from the Paleobiology Database (https://paleobiodb.org), the Ecological Register (http://www.fossilworks.org), the Yale Peabody Museum’s Division of Paleobotany (https://peabody.yale.edu/explore/collections/paleobotany), the Smithsonian National Museum of Natural History (https://collections.nmnh.si.edu/search/paleo), the Japan Paleobiology Database (http://jpaleodb.org), and published literatures. Fossil distribution maps for each genus across different time periods were constructed (Fig. S3). To limit the errors in fossil data, we applied the following criteria to filter the data for unique and most reliable records: (1) The fossil with specific locality information (including fossiliferous deposit, city, or provincial level), (2) The fossil with traceable geological epoch, (3) The fossil is a macrofossil of fruit or leaf, (4) The fossil is associated with a publication, (5) The fossil is assigned to an extant genus or a lower taxonomic rank, and (6) The fossil is accompanied by a detailed description with comparison to extant analogous species. The refined selection criteria yielded a final fossil dataset of 246 records, comprising 61 recognized species in 5 genera, 20 records identified only to the genus level, and 2 records identified only to the section level. The detailed information about the fossils and arguments that support the taxonomy is presented in Tables S8 and S9.

We converted modern coordinates of fossil records to paleo coordinates using the GPLATES WEB SERVICE with the SETON2012 model, according to their geological ages of the collected fossil records (Seton et al., 2012; Müller et al., 2018). Paleoclimate data were collected from multiple Hadley Centre coupled ocean–atmosphere general circulation models (HadCM3L), which incorporate the effects of changes in atmospheric CO² concentration (Beerling and Royer, 2011). Simulations from the Early Cretaceous (ca. 113 Ma) to present were generated using the same underlying algorithms and the same resolution (1° × 1°), including 18 simulations of different periods (Lunt et al., 2016; Farnsworth et al., 2019; Valdes et al., 2021). We used the raster package for R to extract paleoclimatic variables of fossil records based on the results of paleo coordinate conversion and HadCM3L simulations (Hijmans et al., 2015). The extracted paleoclimate data include mean annual temperature (MAT, ℃) and mean annual precipitation (MAP, mm). We calibrated the paleoclimate data prior to the Eocene (including the Eocene) using the climatic parameters obtained via the Coexistence Approach, and excluded obviously anomalous values from the climate simulations (Table S11).

2.2. Calculation of climatic niche position and breadth, and species distribution diversity patterns

To quantitatively assess the niche position and breadth of each species, we performed Principal Component Analysis (PCA) on climate variables for all Ulmaceae species using the FactoMineR package for R (Lê et al., 2008). This approach aimed to simplify multidimensional climate data into a two–dimensional representation of the ecological niche. The selected variables of Bio1, Bio3, Bio9, Bio12, Bio17, and Bio19 were used for the PCA. Firstly, we applied natural logarithmic transformation to non–negative and non–zero variables and standardized all data to minimize the impact of outliers and achieve a normal distribution of niche data.

The first two principal components (PC1 and PC2), which together explained 82.37% of the total climatic variance (PC1: 58.73%; PC2: 23.64%), were used to represent the ecological niche position. The niche position of each species is defined by the median values of PC1 and PC2. Functional richness (FRic) was used to measure niche breadth, representing the area occupied by each species in the two–dimensional principal component space. FRic was performed in TPDs package for R (Carmona et al., 2016). To account for potential data bias due to differences in distribution record scales among species, we computed the niche breadth for temperate clade and tropical clade using the median values of PC1 and PC2 for each species. The Hotelling’s T² test was used to determine whether the differences are significant.

To delineate the study area of each species’ approximate distribution, we used the rangeBuilder package for R to establish the study area for each species (R CoreTeam, 2021). Unlike traditional methods that merely create buffers, rangeBuilder defines polygons based on distribution points and a 25 km radius buffer, preserving the integrity of the study area and preventing fragmentation commonly associated with conventional buffering techniques. Using the cleaned species distribution data and climate variable datasets, we constructed Ecological Niche Models (ENMs) with the Maxent model from the dismo package for R (Hijmans and Elith, 2013).

We randomly divided the climate data into 75% training and 25% testing datasets. The training dataset was utilized to construct ENMs, while the testing dataset evaluated model performance. Based on the established ENMs, we set thresholds using the threshold function in the dismo package for R to derive Species Distribution Models (SDMs). We established the threshold at the point where the sum of sensitivity (true positive rate) and specificity (true negative rate) is maximized. This approach ensures that SDMs predict a greater number of presence grids, as the small geographical scale of grids can lead to distribution fragmentation, and overly stringent thresholds may not accurately reflect reality.

To investigate the diversity patterns and spatial phylogeny of species distributions, we imported the results of SDMs and phylogenetic tree (see Section 2.4 for details) into Biodiverse v.4.3, a tool for the spatial analysis of diversity calculations. These data were used to compute species richness (SR). To investigate the latitudinal diversity patterns of the two clades, we compiled the distribution latitudes of all 58 species within 1–degree latitude ranges.

2.3. DNA extraction, sequencing, and chloroplast genome assembly

The leaf samples were processed by grinding them with a mechanical lapping technique, followed by total DNA extraction using a modified CTAB protocol (mCTAB) (Li et al., 2013). The concentration of the extracted DNA was assessed using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific), while the size of the DNA fragments was evaluated through agarose gel electrophoresis for a selection of samples.

Genome skimming was employed to acquire data from plastid genomes (Dong et al., 2021). Total DNA was fragmented using sonication to produce 350 bp fragments, except for certain herbarium specimens that had degraded to lengths shorter than 350 bp. For these degraded samples, 200 bp insert libraries were constructed, while the other samples were prepared as 350 bp insert libraries using the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA, USA). Sequencing was performed using paired–end reads of 150 bp on the Illumina HiSeq X–ten platform at Novogene in Tianjin, China.

To assemble the chloroplast genome, the raw reads were cleaned and filtered through several steps: Illumina adapter artifacts, low–quality reads, and subpar bases at the ends of the reads were trimmed using Trimmomatic 0.39 with the following parameters: ILLUMINACLIP: TruSeq3–PE.fa: 2:30:10:1: true LEADING: 20 TRAILING: 20 SLIDINGWINDOW: 4:15 (Bolger et al., 2014). Two approaches were used to assemble the chloroplast genome. First, the whole chloroplast genome were assembled using GetOrganelle (Jin et al., 2020), testing a range of k–mer sizes (65, 75, 85, 95, and 105). If GetOrganelle failed to produce complete chloroplast genome, we used the methods following Dong et al. (2022).

Completed chloroplast genome were annotated utilizing the Perl script Plann and PlastidHub (Huang and Cronk, 2015; Zhang et al., 2025), and the missing or incorrect genes were checked in Geneious v.11.1.2. Ultimately, the newly assembled chloroplast genome were deposited in Figshare.

2.4. Phylogenetic analyses and divergence time estimates

Twenty–nine chloroplast genome sequences were downloaded from GenBank and used for subsequent analyses (Table S3). All chloroplast genomes were aligned using MAFFT v.7 (Katoh and Standley, 2013). All alignments were visually checked using Geneious v.11.1.2 and manually adjusted (Kearse et al., 2012). Maximum likelihood (ML) and Bayesian inference (BI) methods were used to infer the phylogenetic relationships of Ulmaceae based on the whole chloroplast genome dataset. The ML tree was generated using RAxML–NG with the best–fit model determined using ModelFinder (Fig. S1) (Kalyaanamoorthy et al., 2017), and branch supports were assessed with 1000 regular bootstrap replicates. BI tree was conducted with MrBayes v.3.2 (Ronquist et al., 2012). Two independent Markov Chain Monte Carlo (MCMC) analyses were performed, each of them runs with three heated and one cold chain for 20 million generations, with sampling for every 100th tree, starting with a random tree. The initial 20% of the samples were discarded as burn–in to confirm the stationarity, and when the average standard deviation of split frequencies was < 0.01, the result was considered to reach stationarity (Fig. S2).

To estimate the divergence time of Ulmaceae, two methods of node–dating and tip–dating were used to estimate the divergence times. For node–dating, a relaxed lognormal clock model in BEAST v.2.6.7 (Bouckaert et al., 2014) was used to infer the divergence times. Seven fossils were used for calibration points (Table S10). The maximum constraint for the root tree was set to 125 Ma (the earliest eudicot fossil pollen), and the maximum constraint was set to 100.5 Ma (the maximum crown age of Cannabaceae) according to the results of Zhang et al. (2021a).

We selected the GTR substitution model and the birth–death speciation process. All fossil calibrations followed Zhang et al. (2021a), using a uniform distribution prior with lower and upper fossil age bounds, with the average of these bounds applied as the age constraint for each calibrated node. Two independent MCMC runs were carried out for 1000 million generations, with samples taken every 10,000 generations. To assess convergence of the Markov chains and the effective sample sizes (ESS) values, the two log files were combined and then checked in Tracer v.1.7.2 (Rambaut et al., 2018). The first 20% of the samples were removed as burn–in in LogCombiner v.2.6.7. The ESS values exceeded 200 for all parameters. The maximum clade credibility (MCC) tree with mean divergence time at the node was generated in TreeAnnotator v.2.6.7 and visualized in FigTree v.1.4.3 (Rambaut, 2015).

For tip–dating, we selected 55 extant species and 61 fossil species from the Ulmaceae dataset. Following the method of Yan et al. (2021), we placed the fossils on the phylogenetic tree. First, we identified the narrowest clade that could be assigned based on the taxonomic placement of the fossil as described in recent publication (for more details, see Table S9). Next, we randomly selected a branch from all branches present prior to the average age of the fossil within that narrowest clade and added the fossil as a species on the chosen branch. For fossils where the specific narrowest clade cannot be determined, they were added as a species to their respective genus based on temporal information. The average age was based on the time range of all specimens of that fossil provided in the publication. The position of the fossil on the branch was determined by the timestamp of the average age. The extinct genus Cedrelospermum was identified as a sister group to the temperate clade (Manchester and Tiffney, 2001).

A relaxed lognormal clock model under the fossilized birth–death (FBD) process model (Heath et al., 2014) in BEAST v.2.6.7 was used to estimate a calibrated phylogenetic tree of extant and extinct species of Ulmaceae (Bouckaert et al., 2014). In the tip dates setting, we refer to the node–dating inference of the divergence time of Ulmaceae. The root age was set as 120 Ma for Ulmaceae, with a lognormal prior distribution (SD = 0.5). Two independent MCMC runs were carried out, and each run 2000 million generations and sampling trees every 10,000 generations. The convergence was examined using Tracer v.1.7.2 (Rambaut et al., 2018) to ensure that the ESS for all parameters was sufficient (i.e., >200). A burn–in of 25% generations was discarded, and TreeAnnotator v.2.6.7 was used to produce a maximum clade credibility (MCC) tree.

2.5. Estimation of diversification rates

To investigate the diversification history of Ulmaceae, 300 trees were randomly sampled from the BEAST tip–dating trees. All duplicate sampled species were removed, and the Lineage–Through–Time (LTT) curves were calculated using the APE package for R (Paradis et al., 2004). To examine diversification dynamics within Ulmaceae, we extracted temperate and tropical clades from each of the BEAST tip–dating trees. Tip labels were matched at the genus level, and the corresponding MRCA subtree was obtained using the get MRCA and extract clade functions in the APE package for R. LTT curves were then generated for all extracted subtrees, and multi–tree LTT plots were used to summarize lineage accumulation across the posterior distribution.

We used an rjMCMC algorithm with the Bayesian analysis of macroevolutionary mixtures (FossilBAMM v.2.6.0) (Mitchell et al., 2019) and BAMMtools package for R to assess the historical diversification dynamics of Ulmaceae (Rabosky et al., 2014) and model rates of speciation, extinction and net diversification. This method allowed branch–specific estimation of diversification rates on non–ultrametric trees (i.e., including fossil taxa) by using a fossilised birth–death process.

The parameter “globalSamplingFraction” was set to 0.87 following the sampling coverage in the current phylogeny to account for incomplete sampling. This parameter could produce unbiased estimates of evolutionary rates under the assumption that species are absent from the phylogeny at random. Prior values were selected using the “setBAMMpriors” function in the BAMMtools package for R. Total MCMC runs were set to 10,000,000 generations and parameters were sampled every 5000 generations. The initial 25% of the samples of the MCMC run were discarded as a burn–in, and convergence was computed using the remaining data. The ESS >200 of the log–likelihood was also determined. Visualization used the BAMMtools package for R.

To better compare the temperate and tropical clades, we used FossilBAMM v.2.6.0 to separately estimate the speciation rate, extinction rate, and net diversification rate for each clade. We set the “globalSamplingFraction” to 0.93 for the temperate clade and 0.58 for the tropical clade. The MCMC was run for 10,000,000 generations, with parameters sampled every 5000 generations. The first 25% of MCMC samples were discarded as burn–in, and convergence was assessed using the remaining samples. In addition, we ensured that the ESS of the log–likelihood exceeded 200. All visualizations were performed using the BAMMtools package for R.

2.6. Species niche variables and reconstructing ancestral niches

Based on the phylogenetic results, we constructed ecological niches of different clades based on the averages of ecological factors. Two environmental variables (MAT and MAP) were selected to reconstruct historical dynamics of the niches. The existing niches of Ulmaceae species were obtained based on the averages of the environmental variables of the grids (Table S12). To reconstruct ancestral niche dynamics, we employed two complementary approaches that differ in their data requirements and analytical scope.

First, ancestral climatic niches were reconstructed using RevBayes (Höhna et al., 2016), based exclusively on the phylogenetic relationships of extant species. This approach does not require fossil trait data and thus allows direct comparison between temperate and tropical clades under a unified analytical framework. Ancestral state estimation was conducted with MCMC analyses run for 10 million iterations. The posterior distributions of ancestral MAT and MAP were summarized and visualized using the ggplot2 package in R (Ginestet, 2011). Differences among clades were assessed using Duncan’s multiple range test implemented in the agricolae package for R.

Second, ancestral niche evolution was inferred using fossilBM (Silvestro et al., 2019), which treats climatic variables as continuous quantitative traits evolving along a time–calibrated phylogeny. Because fossilBM requires reliable and taxonomically assignable macrofossil constraints, this analysis was applied only to lineages with sufficiently dense fossil records, which in Ulmaceae are primarily restricted to the temperate clade. FossilBM analyses were conducted on the tip–dated phylogeny including both extant and fossil taxa. For fossil species represented by multiple occurrences across different localities or time intervals, mean MAT and MAP values were used as trait inputs. Birth–Death MCMC (BDMCMC) analyses were run for five million iterations (Stephens, 1999), sampling every 2500 generations under a Brownian evolutionary model (BM), with the first 10% discarded as burn–in. To quantify temporal changes in niche breadth, trait ranges through time were summarized as 95% credible intervals within one–million–year time bins, based on 100 replicate analyses (Serrano–Serrano et al., 2015). To better visualize how niches changed over the evolutionary history, ggplot2 package for R was used to plot the results (Ginestet, 2011).

To demonstrate the differences in niches between fossil and extant species, we defined biomes based on vegetation types according to Robert Whittaker’s definition, utilizing mean annual temperature and mean annual precipitation data. This analysis was conducted using the plotbiomes package for R.

2.7. Biogeographic history

To fully demonstrate the impact of the fossil record on ancestral area reconstructions and present a more realistic biogeographic history, the node–dated tree and tip–dated tree were used for biogeographic reconstruction. Based on the distribution of extant and fossil species of Ulmaceae and the existence of ocean or plateau barriers between continents, we defined six geographical regions: A) Eastern Asia, B) Indochina, C) Europe, D) Eastern North America, E) The Neotropics, F) Tropical Africa, and G) Western North America (Table S13).

The node–dated phylogenies and tip–dated phylogenies were used the Dispersal–Extinction–Cladogenesis (DEC) model (Ree et al., 2005; Ree and Smith, 2008). To reflect the major changes in continental connectivity, different basic dispersal probabilities among regions were set to reflect inter–region connectivity. The dispersal constraints were divided into four versions (T1: > 100 Ma; T2: 100–30 Ma; T3: 30–5 Ma, and T4: 5–0 Ma, Table S14) according to the history of continental drift and climate change. All analyses were implemented in RASP (reconstruct ancestral state in phylogenies) v.4.3 (Yu et al., 2015). Dispersal, vicariance and extinction events were extracted from the BioGeoBEARS module in RASP.

3. Results 3.1. Diversity distribution patterns and climatic niche divergence of Ulmaceae

The geographic pattern of global species richness (SR) reveals that China contained the highest species diversity globally, comprising 3 genera and 12 species, with major hotspot areas concentrated in the southeastern region. The southeastern United States represents another critical center for SR within the Ulmaceae. Furthermore, high levels of SR are also observed in much of Central and Eastern Europe, Central and South America, south of the Himalayas, and the Indochinese Peninsula.

The temperate clade exhibits significantly higher species diversity than the tropical clade, and the two clades display markedly different latitudinal diversity patterns (Fig. 1b). The temperate clade is concentrated in the Northern Temperate Zone, with only a few extending into tropical regions. The diversity of the temperate clade peaks between 28° N and 38° N in the subtropical zone. In contrast, the tropical clade is nearly entirely restricted to the latitudes between the Tropics of Cancer and Capricorn, with the diversity peak between 5° N and 12° N, predominantly within the tropical zone. The distribution of Ulmaceae species ranges from latitudes 69° N (e.g., Ulmus glabra) to 24° S (e.g., Phyllostylon rhamnoides).

We conducted a principal component analysis on all climatic variables, extracting two components, PC1 and PC2, which explained 58.73% and 23.64% of the climatic niche variation, respectively. The temperate clade and tropical clade showed significant niche differences (T² = 101.66, df = 2.55, P < 0.001, Fig. 1c). Niche breadth was quantified using FRic. In terms of niche breadth, the temperate clade occupied the largest niche space, while the tropical clade had a relatively narrow niche breadth (Fig. 1c), likely due to its considerably fewer species. Additionally, there were differences in niche breadth among species within the temperate clade and tropical clade: most tropical clade species exhibited small niche breadths (FRic < 20), whereas some temperate clade species, such as Ulmus androssowii, Ulmus pumila, and Ulmus lanceifolia, occupied larger niches (FRic > 20) (Fig. 1d).

3.2. Phylogenetic relationships and divergence time of Ulmaceae

Phylogenetic reconstructions using both ML and BI methods strongly support the monophyly of Ulmaceae (ML bootstrap support [BS] = 100, posterior probabilities [PP] = 1; Figs. S1 and S2). Ulmaceae can be divided into two major clades with high support: temperate clade and tropical clade (BS/PP = 100/1; Figs. S1 and S2). Within the temperate clade, Hemiptelea formed the basal lineage and was sister to a clade comprising Ulmus, Planera, and Zelkova (BS/PP = 100/1; Figs. S1 and S2). In this group, Planera and the subgenus Oreoptelea of Ulmus form a highly supported clade (BS/PP = 100/1; Figs. S1 and S2), which was sister to Zelkova (BS/PP = 92/1; Figs. S1 and S2). Within the tropical clade, Ampelocera and Phyllostylon formed a clade that was sister to Holoptelea (BS/PP = 100/1; Figs. S1 and S2).

Within Ulmus, two subgenera were recognized: Oreoptelea (BS/PP = 100/1) and Ulmus (BS/PP = 100/1), and the genus was further divided into six sections: Sect. Ulmus, Sect. Microptelea, Sect. Lanceaefoliae, Sect. Blepharocarpus, Sect. Chaetoptelea, and Sect. Trichoptelea. In the subgenus Ulmus, Sect. Microptelea and Sect. Lanceaefoliae were resolved as a clade (BS/PP = 100/1), formed a clade that was sister to Sect. Ulmus (BS/PP = 100/1), and this combined clade was sister to Ulmus villosa (BS/PP = 100/1). Within subgenus Oreoptelea, Sect. Blepharocarpus and Sect. Chaetoptelea clustered together (BS/PP = 67/0.79), which was sister to the clade formed by Sect. Trichoptelea + Ulmus thomasii clade (BS/PP = 100/1), and this larger clade was sister to Ulmus elongata (BS/PP = 100/1).

Several conflicts of intrageneric relationships were found between the ML and BI methods. In the ML tree (Fig. S1), Ulmus rubra and Ulmus serotina were resolved as a clade (BS = 99), which was sister to Ulmus americana (BS = 100) within Sect. Blepharocarpus. In contrast, in the BI tree (Fig. S2), Ulmus rubra and Ulmus americana were resolved as sister species within Sect. Blepharocarpus (PP = 1), while Ulmus serotina and Ulmus crassifolia were sister species within Sect. Trichoptelea (PP = 1).

The tip–dating results indicated that the stem age of Ulmaceae was at the Early Cretaceous (Fig. S4 and Table S15). The crown age of Ulmaceae, marking the divergence of temperate and tropical clades, was dated to the Late Cretaceous (mean: 85.21 Ma, 95% the highest posterior density [HPD]: 83.24–87.16 Ma). Within the temperate clade, Cedrelospermum diverged from other temperate genera ca. 80.28 Ma (95% [HPD]: 72.46–86.04 Ma), while Hemiptelea diverged from other extant temperate genera ca. 75.8 Ma (95% [HPD]: 68.56–82.82 Ma). Stem and crown age of Zelkova were ca. 69.72 Ma (95% [HPD]: 63.13–76.96 Ma) and ca. 54.11 Ma (95% [HPD]: 52.19–56.06 Ma). The crown age of Ulmus and Planera was ca. 67.99 Ma (95% [HPD]: 62.6–74.55 Ma). In the tropical clade, Holoptelea diverged from Ampelocera and Phyllostylon at ca. 52.21 Ma (95% [HPD]: 32.32–74.02 Ma) and the divergence time between Ampelocera and Phyllostylon was inferred to ca. 34.86 Ma (95% [HPD]: 17.29–53.67 Ma). The origin times of most clades were during the Early Eocene to the Late Miocene (ca. 55.83–5.55 Ma). This result shows a marked discrepancy compared to the inference derived exclusively from extant taxa (Fig. S5).

3.3. Diversification rates of Ulmaceae

The Lineage–Through–Time (LTT) plot shows that Ulmaceae exhibited a gradual accumulation of lineage diversity from the Late Cretaceous to the Early Eocene (before 50 Ma). From the Middle to Late Eocene (47.8–37.8 Ma), diversity clearly decreased, and during the interval from the Late Eocene to the Early Oligocene (37.8–33.9 Ma), the slope of the LTT curve became nearly flat, indicating a marked slowdown in diversification. After the Early Miocene, lineage diversity increased rapidly (Fig. S6a). The temperate and tropical lineages displayed notably different diversification dynamics. For the temperate lineage (Fig. S6b), the pattern of lineage accumulation closely resembled that of the family as a whole. Prior to 50 Ma, lineage growth was slow, gradually increasing to approximately 20 lineages. Between 47.8 and 33.9 Ma, the curve became slightly flattened, reflecting a temporary reduction in accumulation rate. Following the onset of the Early Miocene, the curve exhibited a sharp increase, with lineage numbers rising rapidly to more than 40 at present. In contrast, the tropical lineage showed a generally flatter LTT curve during its early evolutionary history. Lineages accumulated gradually through the Middle Eocene, and the fastest increase in lineage accumulation occurred during the Miocene (Fig. S6c).

The BAMM results indicate a significant divergence in diversification rates between the temperate and tropical clades during the early stages of evolution in the Late Cretaceous. The temperate clade experienced a marked acceleration in diversification rate, which was substantially higher than that of the tropical clade (Fig. 2a). Both the speciation and extinction rates (Fig. 2b) indicated an initial increase in the temperate clade, followed by a period of overall stability ca. 80 Ma, with a slight subsequent decline. In contrast, the tropical clade exhibited a rapid decline in the early speciation rate, stabilizing ca. 80 Ma. Overall, the Ulmaceae experienced a decline in early speciation rates, followed by an upward trend, a subsequent decline during the Miocene, and then stabilization. The net diversification rates remained remarkably stable across time in Ulmaceae and both clades (Figs. S7 and S8). The temperate clade exhibited consistently low net diversification rates, with values fluctuating only minimally around zero from the Late Cretaceous to the present. The tropical clade showed slightly higher net diversification rates overall, but its trajectory was similarly stable, without any indication of early accelerations or subsequent slowdowns. These results suggest that neither clade experienced pronounced temporal variation in diversification dynamics. Rather than being driven by discrete diversification bursts, the long–term evolutionary history of Ulmaceae appears to reflect relatively constant diversification regimes within both the temperate and tropical lineages.

3.4. Reconstruction of ancestral niches

The results from RevBayes show that the mean annual temperature (MAT) for extant species in the temperate clade is approximately 9.75 ℃, while their ancestors thrived in a warmer environment with a temperature of about 14.15 ℃. In contrast, the average temperature for extant species in the tropical clade is approximately 24.96 ℃, which is similar to that of their ancestors (MAT: about 24.32 ℃) (Fig. 3a and Table S16). In terms of precipitation, the mean annual precipitation (MAP) for extant species in the temperate clade is approximately 898.33 mm, whereas their ancestors lived in a more humid environment with a mean annual precipitation of about 1157.47 mm. On the contrary, extant species in the tropical clade inhabit a relatively humid environment with a MAP of approximately 1800.64 mm, while their ancestors were adapted to a drier environment (MAP: about 1268.79 mm) (Fig. 3b and Table S16). Additionally, the niche breadth of the temperate clade expanded significantly during the evolutionary process, which may indicate that the temperate branch as a whole has undergone cold tolerance adaptation in its evolutionary history (Fig. 3a). During the early evolution of Ulmaceae, the differentiation between temperate and tropical clades was accompanied by significant niche differentiation. Moreover, the results from FossilBM indicate that both MAT and MAP of the temperate groups maintained relatively constant evolution, suggesting that they retained similar temperature and precipitation preferences throughout the evolutionary process (Fig. 3c and d). Furthermore, to demonstrate the differences in niches between fossil and extant species, we utilized MAT and MAP to analyze the Whittaker biogeographic community types of various species. The results indicated that extant and fossil species cluster together, showing no consistent trend along the time scale, thus further supporting the notion of niche conservatism of the temperate clade (Fig. 3e). Extant and fossil species in the temperate clade are distributed across temperate and subtropical regions, occupying large and diverse niches. Species in the tropical clade are restricted to tropical rainforest and tropical seasonal forest ecosystems.

3.5. Biogeographic history without fossil taxa

The ancestral area reconstruction based on node–dating of the phylogenetic tree supported the East Asian origin of the Ulmaceae (Fig. 4). There was a total of 26 dispersal events, 15 vicariance events, and 4 extinction events. The temperate clade initially diversified in East Asia, subsequently dispersing to eastern North America during the Late Cretaceous, which led to the ancestor of Ulmus, Planera, and Zelkova.

The Ulmus originated in East Asia or eastern North America. During the Late Cretaceous (ca. 77.19 Ma), the Subgenus Oreoptelea diverged in eastern North America. Subsequently, it dispersed to East Asia via the Bering Land Bridge (ca. 75.69 Ma) and then spread from eastern North America to Europe across the North Atlantic land bridge (ca. 74.35 Ma). This subgenus later re–dispersed from Europe back to eastern North America (ca. 73.00 Ma). Additionally, at ca. 63.09 Ma, it dispersed from eastern North America to the Neotropics. The subg. Ulmus diverged in the Late Cretaceous (ca. 71.31 Ma) in the Indochina and subsequently dispersed to East Asia and Europe, leading to the ancestor of sect. Lanceaefoliae, sect. Microptelea, and sect. Ulmus. In the Early Paleocene (ca. 64.19 Ma), it spread from East Asia to the Indochina and Europe.

Zelkova diverged in the Late Cretaceous (ca. 79.18 Ma) between East Asia and Europe, followed by a vicariance event between these regions, which resulted in the current distribution pattern. Hemiptelea and Planera originated in East Asia and eastern North America, respectively, without subsequent dispersal events. The extant species are currently distributed in East Asia and eastern North America.

The tropical clade diverged in Europe or eastern North America, followed by three dispersal events that led to its spread to the Indochinese Peninsula, Neotropical regions, and Africa, ultimately resulting in the current distribution pattern of the tropical clade.

3.6. Biogeographic reconstruction including fossil taxa and dynamic history of the distribution pattern

Tip–dating analysis using the DEC model, which incorporated 61 fossil occurrences and 55 extant species, supported a north temperate origin for Ulmaceae during the Late Cretaceous (relative probability = 0.34). The temperate clade most likely originated from western North America (ca. 80.28 Ma) and dispersed to East Asia, giving rise to the ancestors of all extant genera of the temperate clade (Fig. 4).

Overall, the biogeographic reconstruction inferred 78 dispersal events, 34 vicariance events, and 7 extinction events across the evolutionary history of Ulmaceae. Among these, 62 were intercontinental dispersal events, predominantly between North America and Asia, and between North America and Europe (Fig. 5b and Table S17). When temporal context is considered, East Asia acted as a persistent dispersal source throughout most of the evolutionary history, whereas western North America functioned as both a major source and a sink prior to the Oligocene. Europe primarily acted as a sink, particularly after the Oligocene (Fig. 5b).

The earliest dispersal occurred in the Late Cretaceous (ca. 85.21 Ma). In the Early Eocene, there was a dispersal peak, with western North America serving as the primary hub (12 emigrations and 10 immigrations). During this period, species diversity in western North America grew most rapidly, reaching its zenith, while East Asia maintained stable diversity and Europe showed continuous increases (Fig. 5c). In the Middle–Late Eocene, the dispersal events declined significantly. Species diversity in western North America began a steep decline, stabilized in eastern North America, and rose to its peak in Europe. A second dispersal peak occurred during the Oligocene, whereas western North America transitioned to a balanced hub (5 emigrations/immigrations), and Europe received 10 immigrations. Concurrently, East Asian diversity surged to its peak, western North America’s decline slowed, and Europe and eastern North America exhibited minor diversity fluctuations. After the Miocene, the dispersal events sharply decreased (10 events). Species diversity in western North America collapsed to zero, East Asia and Europe experienced slight declines, and eastern North America briefly rebounded before all regions entered sustained decline (Fig. 5c).

At the genus level, individual lineages broadly followed these temporal and regional patterns (Fig. 5a and Fig. S9). Ulmus experienced extensive intercontinental dispersal during the Paleogene. The subg. Ulmus originated in East Asia and dispersed to Europe and the Indochina during the Late Paleocene, followed by bidirectional exchanges between East Asia and western North America in the Late Paleocene–Early Eocene. Subsequent dispersal into Europe during the Oligocene was accompanied by regional extinction in North America, resulting in the present–day East Asia–Europe disjunction. In contrast, the ancestors of subg. Oreoptelea and Planera originated in western North America, with repeated exchanges between western and eastern North America during the Early Eocene. Planera subsequently became restricted to eastern North America, whereas subg. Oreoptelea later dispersed to Europe in the Late Eocene and reached the Neotropics during the Late Miocene.

Zelkova originated in East Asia and dispersed repeatedly to western North America from the Early Eocene through the Late Oligocene, but became extinct in North America during the Miocene. Independent dispersal events from East Asia to Europe during the Oligocene and Early Miocene gave rise to the extant European species. Hemiptelea, by contrast, showed no evidence of long–distance dispersal and has remained confined to East Asia throughout its evolutionary history. The extinct genus Cedrelospermum originated in western North America and dispersed to eastern North America and Europe during the Early Eocene. During the Oligocene, there were further dispersal events from East Asia to the Neotropics and Europe, before becoming globally extinct after the Miocene.

The ancestor of the tropical clade dispersed from Europe to the Indochina and the Neotropics during the Early Eocene (ca. 52.21 Ma). Within the Neotropics, the divergence formed Phyllostylon (including Ampelocera), while the divergence in the Indochina led to the establishment of Holoptelea. Subsequently, Holoptelea dispersed from the Indochina to Africa during the Middle Miocene (ca. 15.35 Ma). The discontinuous distribution of the tropical clade in the Southern Hemisphere was established at least by the Middle Miocene (Fig. 5a).

4. Discussion 4.1. Species richness anomalies between temperate and tropical clades

Our fossil–integrated analyses indicate that, although both the temperate and tropical clades of Ulmaceae show relatively stable net diversification rates through time, their extant species richness differs markedly. This pattern implies that the observed disparity in species richness is unlikely to be driven primarily by major shifts in speciation or extinction rates (Tietje et al., 2022). A more parsimonious explanation is that differences in clade diversification times and the ecological context of diversification (consistent with time–dependent effects and density–dependent constraints) underlie the contrasting richness patterns (Rabosky, 2009; Wiens, 2011; Henao Diaz et al., 2019).

Time–dependent effects are likely a major factor underlying the difference in species richness (Ulrich, 2006; Marin and Hedges, 2016). The temperate clade diversified earlier and therefore had a much longer time span over which species could accumulate. When net diversification rates remain relatively constant, older clades simply have more time for lineage accumulation and thus tend to exhibit higher species richness (McPeek and Brown, 2007; Ho et al., 2011). In other words, the early divergence of the temperate clade provided a “first–mover” advantage for species accumulation. These time–dependent effects have been widely regarded as a major determinant of differences in species diversity among clades and regions (Stephens and Wiens, 2003; Wiens, 2011; Henao Diaz et al., 2019). Previous studies have shown that long–term diversity patterns can arise under nearly constant diversification rates, without requiring rate accelerations (Nee et al., 1992; Ricklefs, 2007; Rabosky, 2009). In this context, the higher species richness of the temperate clade may reflect its earlier diversification and longer history of lineage accumulation, rather than intrinsically higher diversification rates.

In contrast, the tropical clade diversified later and did not undergo substantial expansion until the Miocene (Fig. S6c). Because its diversification history spans a shorter time interval, the tropical clade has had fewer opportunities for species accumulation. These time–dependent effects help explain why the temperate clade currently contains 46 species, whereas the tropical clade includes only 12 species, even though their long–term diversification rates appear similar. In addition, the tropical clade appears to have colonized already species–rich tropical communities during its diversification. Under such conditions, diversification may have been constrained by density–dependent processes, including ecological saturation and intensified interspecific competition (Walker and Valentine, 1984; Valentine, 1985; Rabosky and Lovette, 2008). When a lineage colonizes an already diverse ecosystem, new species may face strong competition and limited unoccupied niche space, making diversity accumulation more difficult and thereby restricting further adaptive radiation (Etienne et al., 2012; Rabosky, 2013; Condamine et al., 2019). Consequently, the relatively low species richness of the tropical clade, despite its stable diversification rates, may reflect the combined effects of time–dependent accumulation and density–dependent limits.

Overall, the net diversification rates of both clades remained relatively stable through time, with no clear evidence of major accelerations or slowdowns. This pattern suggests that long–term differences in species diversity are more likely driven by variation in lineage age and ecological constraints, rather than by episodic shifts in diversification rates (Ricklefs, 2007; Wiens, 2011). Therefore, the temperate clade’s earlier diversification and longer accumulation history, together with the tropical clade’s later diversification and potential density–dependent constraints, jointly account for the striking contrast in species diversity observed within Ulmaceae.

4.2. Climatic niche evolution drives contrasting diversity and distribution between temperate and tropical clades

The temperate and tropical clades of Ulmaceae exhibit markedly different species richness and biogeographic distributions (Fig. 1a), which are also closely associated with contrasting patterns of climatic niche evolution. Our ancestral niche reconstructions indicate that the temperate clade shows pronounced niche conservatism, maintaining relatively stable mean annual temperature (MAT) and precipitation (MAP) through time, while simultaneously exhibiting a broader niche breadth and a gradually expanding climatic tolerance range (Fig. 3). This combination allowed temperate lineages to respond to long–term climatic fluctuations primarily through geographic migration and regional replacement, rather than through substantial shifts in their climatic preferences, particularly during major cooling events such as the Eocene–Oligocene Transition (Wiens and Graham, 2005).

Although global mean temperatures during the Late Cretaceous remained high, increasing climatic heterogeneity, latitudinal thermal gradients, and episodic cooling trends toward the end of the Cretaceous began to restructure terrestrial ecosystems. The Late Cretaceous ‘greenhouse’ world witnessed a transition from one of the warmest climates of the past 140 million years to cooler conditions (Hu and Wang, 1999; Miller et al., 2005; Thibault and Gardin, 2006; Friedrich et al., 2012; Linnert et al., 2014).

This climatic context created a more suitable living environment for the temperate clade, enabling it to quickly adapt to the cooling trend. In contrast, the tropical clade, which relies on warm environments, faced significantly increased survival pressure during the cooling period. Thus, the first major divergence between the two clades likely reflects differences in ecological tolerance and geographic opportunity, rather than pronounced shifts in diversification rates. Subsequently, during multiple global cooling events such as the Eocene–Oligocene Transition and the Miocene (Zachos et al., 2001), the temperate clade further leveraged its geographic migration strategy to successfully expand its distribution range to the mid–to–high latitudes of the Northern Hemisphere. In this process, its distribution pattern was continuously consolidated, further solidifying its dominant status in temperate ecosystems (Manchester and Tiffney, 2001). Consistent with this pattern, our tip–dating results indicate that all extant genera of Ulmaceae had diverged before the Middle Miocene, with most temperate species diversifying from the Early Eocene to Late Miocene.

In contrast, the tropical clade exhibited restricted niche evolution, with both ancestral and extant species favoring warm, humid conditions and showing limited climatic adaptability. This ecological constraint confined tropical species to low–latitude tropical regions and limited their capacity for migration and replacement events, resulting in its lower present–day species richness.

Nevertheless, due to the lack of reliable fossil records for the tropical lineage, it remains difficult to determine whether the origin of the tropical clade reflects a true ecological niche shift from cooler and drier northern environments, or whether warm and humid niche preferences were already present prior to its dispersal into tropical regions (Kidwell and Holland, 2002). Based on the available evidence, this lineage appears to have expanded into the tropics shortly after a single major diversification event separating it from the temperate clade. However, the mechanisms underlying this transition—whether driven by the transient Paleogene warming events, ecological filtering, or pre–existing adaptations to warm environments—remain unresolved (Kidwell and Holland, 2002). Future discoveries of well–preserved tropical fossils or independent paleoclimatic proxies will be crucial for constraining the timing and environmental context of tropical dispersal and niche evolution in this clade (Blois et al., 2025).

4.3. Intercontinental diversity in Ulmaceae

In addition to climatic niche dynamics, differences in fruit morphology and dispersal mechanisms further reinforced the contrasting biogeographic trajectories of temperate and tropical clades in Ulmaceae. Within the temperate clade, genera such as Ulmus and Hemiptelea possess winged fruits adapted for wind dispersal, a trait that likely facilitated long–distance dispersal and repeated intercontinental exchanges in the Northern Hemisphere (Todzia, 1993). In contrast, the asymmetric, unwinged drupes of Zelkova are considered a relatively primitive fruit type; mature fruits commonly fall together with the entire twig, while the attached dried leaves function as drag–enhancing structures that reduce impact upon landing, analogous to a parachute mechanism (Oyama et al., 2018; Certini et al., 2020). Planera produces fruits with fleshy protuberances and typically occurs along river systems, suggesting adaptation to water–mediated dispersal (Todzia, 1993). In the Neotropics, Ampelocera bears relatively large and often brightly colored drupes that are primarily dispersed by birds (Todzia, 1989). These non–wind–dispersal strategies may impose limits on dispersal distance or introduce ecological specificity, potentially reducing opportunities for large–scale range expansion (Seidler and Plotkin, 2006; Huang et al., 2024). Moreover, although some tropical genera (Holoptelea and Phyllostylon) possess wind–dispersed fruits, extensive oceanic barriers may have limited the effectiveness of wind–mediated intercontinental dispersal in tropical regions (Larrue et al., 2023).

Fossil–integrated biogeographic reconstructions indicate that the temperate clade of Ulmaceae originated in western North America and rapidly expanded across mid–to–high latitudes of the Northern Hemisphere during the Paleocene and Early Eocene (Fig. 5). This early expansion was facilitated by major intercontinental dispersal routes. Boreal migrations between North America and East Asia likely occurred via the Bering Land Bridge, active from the Paleogene to the Late Neogene (Tiffney and Manchester, 2001). Trans–Atlantic dispersals between North America and Europe were likely enabled by the North Atlantic Land Bridge, which persisted as an island–stepping corridor until the Late Miocene (Denk et al., 2010, 2011). Additional floristic exchange between East Asia and Europe may have followed the Paleocene closure and the Eocene retreat of the Turgai Strait (Tiffney and Manchester, 2001). These dispersal routes were particularly effective during the Paleocene–Eocene Thermal Maximum (PETM), when reduced latitudinal temperature gradients and globally warm, equable climates promoted poleward range expansion and frequent intercontinental migrations (Wing et al., 2005; Jaramillo et al., 2010; Korasidis et al., 2022). Consistent with this pattern, western North America functioned as a major biogeographic hub during the Early Eocene, experiencing numerous emigration and immigration events that contributed to rapid lineage accumulation and diversification (Wolfe, 1975; Tiffney, 1985a, 1985b; Manchester, 1999; Jones et al., 2013).

The Eocene–Oligocene Transition (EOT) marked a major turning point in Ulmaceae biogeography. Abrupt global cooling, increased temperature seasonality, and declining atmospheric CO₂ were accompanied by a pronounced reduction in precipitation and effective moisture, particularly in western North America (Retallack, 2007; Pearson et al., 2009; Pound and Salzmann, 2017). These climatic changes, promoted regional aridification and the fragmentation of mesic temperate forests (Terry Jr, 2001; Sheldon et al., 2012; Boardman and Secord, 2013). As a consequence, dispersal corridors deteriorated, range contractions and sharply reduced intercontinental dispersal from western North America during and after the EOT. In contrast, East Asia emerged as a primary refugial region and diversification center during the EOT. Relatively buffered climatic conditions and complex topography facilitated lineage survival and promoted regional diversification (Tang et al., 2018). While global cooling imposed strong selective pressures, East Asia retained sufficient ecological continuity to support the maintenance and initial radiation of the temperate clade, setting the stage for subsequent diversity accumulation (López–Pujol et al., 2011).

Following the Miocene, the progressive deterioration of intercontinental land bridges, together with continued global cooling, likely imposed increasing constraints on long–distance dispersal across the Northern Hemisphere (Zachos et al., 2001). As dispersal opportunities became more limited, regions characterized by relatively stable climates and high topographic complexity—particularly East Asia—may have shifted to a species diversification pattern driven by increased in–situ diversification (Kubota et al., 2017; Tian et al., 2024). Under such conditions, rapid species accumulation could have been driven by a combination of prolonged lineage persistence and also has highly heterogeneous environments because of its topographic complexity. In contrast, western North America and northern Eurasia experienced repeated range contractions and local extirpations, whereas refugial regions in East Asia, southeastern North America, and southwestern Eurasia preserved relict lineages, ultimately establishing the modern disjunct distribution pattern of Ulmaceae (Milne, 2006).

Collectively, these temporally structured dispersal, extinction, and diversification processes—mediated by climatic change, tectonics, and lineage–specific dispersal capacity—governed the long–term persistence and extinction of Ulmaceae, ultimately shaping their present–day intercontinental distribution patterns.

5. Conclusions

By integrating a fossil–informed phylogenomic framework, this study reveals the deep–time processes underlying the asymmetric diversity and distribution patterns of Ulmaceae. Our results indicate that the family likely originated in temperate regions of the Northern Hemisphere during the Late Cretaceous and experienced extensive intercontinental dispersal and range expansion throughout the Paleogene.

The contrasting present–day diversity of the temperate and tropical clades is best explained by early divergence in climatic niche evolution coupled with differing biogeographic opportunities. Temperate lineages retained strong niche conservatism while expanding their niche breadth, facilitating repeated migration across high–latitude land bridges and promoting long–term lineage persistence. In contrast, tropical lineages exhibit more restricted niche evolution and limited geographic expansion, resulting in lower species richness. Notably, these long–term diversity differences are not associated with major temporal shifts in diversification rates, which remained relatively stable in both clades.

Overall, our findings underscore the critical role of fossil integration in reconstructing biogeographic history, as fossils refine estimates of divergence times, ancestral ranges, and evolutionary trajectories that cannot be recovered from extant taxa alone. This fossil–integrated perspective provides a robust framework for understanding how climatic niche evolution and deep–time biogeographic processes jointly shape modern plant diversity.

Acknowledgements

This work was supported by the Discipline Crossing Foundation of School of Ecology and Nature Conservation, Beijing Forestry University (BH2025–JX–01), Fundamental Research Funds for the Central Universities (ZZK202502) and Science and Technology Basic Resources Investigation Program of China (Grant No. 2021FY100200).

Data availability

The datasets generated and analyzed during this study are available in the supplementary materials and have been deposited in Figshare at https://doi.org/10.6084/m9.figshare.31675672.

CRediT authorship contribution statement

Xin Yan: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Formal analysis, Data curation, Conceptualization. Enze Li: Software, Formal analysis, Data curation, Methodology. Xingyong Cui: Conceptualization, Methodology, Formal analysis. Liangcheng Zhao: Validation, Supervision, Funding acquisition. Wenpan Dong: Writing – review & editing, Writing – original draft, Conceptualization, Funding acquisition, Supervision, Validation.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2026.03.011.