The past, present and future distribution of Sargentodoxa Rehder & E.H.Wilson: Perspectives from fossil record and species distribution models

The past, present and future distribution of Sargentodoxa Rehder & E.H.Wilson: Perspectives from fossil record and species distribution models | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article The past, present and future distribution of Sargentodoxa Rehder & E.H.Wilson: Perspectives from fossil record and species distribution models Xuanqi Liu, Huasheng Huang, Xia Meng, Minqiao Li, Zeyu Qin This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6516979/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Global climate change is a critical factor influencing biodiversity and ecosystem stability by altering the suitable habitats of many species. Sargentodoxa cuneata , is an endemic and relict plant species in China. Identifying its suitable habitats across different periods and glacial refugia helps explain how S. cuneata survived Quaternary climate fluctuations, which is crucial for informing its future conservation. However, long-term tracking of its distribution and systematic description of biogeographical evolution remains scarce. Here, we compare ten species distribution models to assess their predictive performance. Ultimately, we apply a random forest model to simulate the suitable habitats of S. cuneata under past, present, and future climate scenarios and integrate fossil records to analyze its biogeographical history. We find that S. cuneata is currently distributed primarily south of the Qinling-Huaihe Line in China, particularly in mid- and low-altitude mountainous regions with abundant precipitation and moderate temperatures. During the Last Glacial Maximum (LGM, about 22,000 years ago) and Mid-Holocene (MH, about 6,000 years ago), its suitable habitat contracted significantly, with extremely suitable areas nearly disappearing due to colder climate. Glacial refugia are identified in three mountain ranges within Central and South China. Model simulations under two different climate scenarios suggest that while the total suitable habitat of S. cuneata may expand, extremely suitable areas could decline, with a northward expansion and southern contraction. This study will provide insights into the long-term impact of climate change on relict plant species and contribute to a better understanding of the evolutionary history of East Asian flora. Lardizabalaceae Habitat suitability Glacial refugia Random Forest model Machine learning Biogeography Climate change Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 1 Introduction Over the past few decades, the global average temperature has steadily risen, accompanied by an increasing frequency of extreme climatic events. This have significantly impacted ecosystem stability and species adaptive capacities. Since the Quaternary period, glacial-interglacial cycles have driven habitat shifts in many extant species across the Northern Hemisphere [1], and also led to genomic divergence [2, 3]. Climate change not only reshapes the habitat suitability but also contributes to habitat loss, population declines, and even extinction [4]. In the face of escalating climate change, species geographical distribution patterns are undergoing profound transformations. Accurately predicting these distributional shifts and glacial refugia is essential for biodiversity conservation, biological resource management, and sustainable development [5]. The species distribution models (SDMs), also known as ecological (or environmental) niche models (ENMs), have become a fundamental tool for assessing the influence of environmental factors on species distributions. The SDMs have been widely applied in tracing ecological indication and biogeographical history [6–9]. They estimate potential suitable habitats for species across different temporal scales by modeling the relationship between known species occurrences and environmental variables [10]. The SDM algorithms that have been commonly used include maximum entropy model (MaxEnt) [11], random forest (RF) [12], generalized linear models (GLM) [13], and generalized additive models (GAM) [14], among others. The selection of the most appropriate modeling approach depends on the characteristics and objectives of the study [15, 16]. Glacial refugia refer to regions that provided climatically suitable conditions for species persistence during global glacial periods, such as the Last Glacial Maximum (LGM) [17]. These areas are sometimes also referred to as climate refugia [18]. Plants, highly sensitive to climatic fluctuations, rely on refugia not only for survival but also as centers of genetic differentiation and adaptive evolution. Consequently, glacial refugia play a critical role in shaping evolutionary processes, maintaining biodiversity, and influencing present-day species distribution patterns [19]. During glacial periods, many species experienced severe range contractions or even local extinctions due to temperature declines and ice sheet expansion. However, certain regions with complex topography and diverse microclimates, such as Southwest China [19], the Mediterranean Basin [20] and the Himalayas [21], served as crucial refugia for subtropical and temperate plants. These areas not only facilitated species survival during glacial episodes but also acted as postglacial centers of recolonization, shaping contemporary distribution patterns. By using species distribution models (SDMs), we can delineate the current suitable habitat range of a species, reconstruct its historical distribution patterns under past climatic conditions, and thus identify its glacial refugia. Additionally, the SDMs can be used to predict potential habitat shifts under future climate change scenarios, and this provides evidences for the development of effective conservation strategies. The genus Sargentodoxa Rehder & E.H.Wilson belongs to the family Lardizabalaceae, and is a monotypic genus containing only Sargentodoxa cuneata (Oliv.) Rehd. et Wils (Fig. 1). S. cuneata is a relict species endemic to China, and primarily distributed in subtropical regions, with occasional occurrences in northern Laos and Vietnam. It commonly grows in well-lit open forests, forest edges, and shrublands on mountain slopes at elevations of 100–1000 m [22]. S. cuneata holds significant medicinal value [23]. Its bioactive compounds exhibit anti-inflammatory, antibacterial, and antitumor properties, which are helpful for treating various diseases, particularly arthritis and sepsis [24]. Consequently, the sustainable utilization and conservation of this species have attracted considerable attention. Currently, research on medicinal properties of S. cuneata is relatively abundant [25, 26], while studies on its spatiotemporal distribution remain limited [27]. Climate change is expected to have profound impacts on the spatiotemporal distribution of S. cuneata . Fossil evidence indicates that S. cuneata has once broader distribution (across the globe) over the past several million years than at present. However, following large-scale extinctions in the Americas and Europe, its habitat is now restricted to certain regions of East Asia [28, 29]. The identification of historical glacial refugia for S. cuneata can shed light on why its current distribution is limited to China and offer insights into its potential adaptability to different climatic conditions. Furthermore, these refugial regions may still serve as contemporary genetic diversity hotspots for S. cuneata . Recognizing and protecting these areas is crucial for preserving genetic resources and enhancing the species resilience to future climate change. Although Sargentodoxa cuneata is a relict species of significant ecological and medicinal value, studies on its suitable habitat distribution remain scarce. Therefore, this study aims to address the following key questions: How does the suitable habitat of S. cuneata change from the past to present and the future in response to climate change? Where were the glacial refugia for S. cuneata during the ice age? To answer these questions, we apply species distribution models (SDMs) to simulate the suitable habitat distribution of S. cuneata under past, present, and future climatic scenarios. We analyze changes in suitable habitat and the species responses to climate change. We also integrate fossil evidence to untangle its biogeographical history. We hope this study will provide valuable insights into the biogeographical history of S. cuneata , and shed light on maintaining the stability of East Asian ecosystems and promoting biodiversity conservation under the scenario of global climate change. 2 Materials and Methods 2.1 Data collection The modern distribution data of Sargentodoxa cuneata were obtained from the Global Biodiversity Information Facility (http://www.gbif.org, GBIF.org, 2024) and the National Plant Specimen Resource Center (CVH, http://www.cvh.ac.cn/). To ensure data quality, outliers and duplicate records were removed using CoordinateCleaner [31] and ENMTools packages [32] in R. This ensures that a maximum of one occurrence point was retained per 5-minute grid cell. Additionally, two anomalous occurrence points from Sri Lanka and northeastern China were identified. After verification with Flora Reipublicae Popularis Sinicae (FRPS, https://www.iplant.cn/) and Plants of the World Online (POWO, https://powo.science.kew.org/), these two records were deemed erroneous and subsequently removed. This resulted in a total of 794 geographically valid occurrence records for ecological niche modeling (Fig. 2). To facilitate robust species distribution modeling, 2,400 pseudo-absence points were generated. Using the disk method in the biomod2 package [33], pseudo-absence points were selected within a circular buffer around presence points, with a minimum distance of 10 km. Fossil records of the relict species Sargentodoxa cuneata play a crucial role in identifying its potential past refugia. A total of 10 fossil records were retrieved from the literature, including nine macrofossil records (e.g., Manchester, 1999) primarily consisting of seeds, and one Chinese pollen record [35]. These fossil data were used to validate and disentangle the historical distribution patterns of S. cuneata (Table 1). Table 1. Fossil records of Sargentodoxa cuneata . Record ID Organ type Form-species Latitude Longitude Country Age Reference 1 Seed S. globosa Manchester 44.66 -120.30 USA Middle Eocene Manchester, 1999 2 Seed S. lusatica 51.30 12.40 Germany Late Eocene–Late Oligocene Mai, 2001 3 Fruit/seed S. cuneata 43.83 -73.05 USA Early Miocene (mid-Tertiary) Tiffney, 1993 4 Seed S. sp. 31.35 -89.30 USA Middle Miocene McNair et al., 2019 5 Seed S. lusatica 51.50 14.20 Germany Middle Miocene Mai, 2001 6 Macrofossil S. cuneata 35.30 137.10 Japan 10 Ma Momohara, 2001 7 Macrofossil Sargentodoxa 36.21 -82.38 USA Late Miocene–Early Pliocene Mead et al., 2012 8 Seed Sargentodoxa 48.58 7.75 French Late Miocene–Early Pliocene Geissert et al., 1990 9 Seed Sargentodoxa 42.00 13.00 Italy Pliocene Martinetto, 2001 10 Pollen S. cuneata 27.34 103.73 China Pliocene–Pleistocene Song, 1988 2.2 Environmental variables We selected 19 bioclimatic variables that could potentially influence the distribution of Sargentodoxa cuneata from the WorldClim Database version 2.1 (www.worldclim.org), with a spatial resolution of 5 minutes. To prevent model overfitting due to multicollinearity among environmental variables, we calculated the correlation coefficients between variables (Fig. 3). Variables showing a correlation coefficient with an absolute value less than 0.8 (|r| < 0.8) were retained, and this results in the selection of nine bioclimatic variables. Additionally, elevation data were sourced from WorldClim, and topographic factors including slope and aspect were calculated using ArcGIS. This leads to a final set of 12 environmental variables for modeling (Table 2). Table 2. Environmental factors for modeling SDMs. Variables Description Units Bio1 Annual mean Temperature ℃ Bio2 Mean diurnal range (mean of monthly (max temp-min temp)) ℃ Bio3 Isothermality (bio2/bio7) (×100) / Bio7 Temperature annual range (bio5-bio6) ℃ Bio13 Precipitation of wettest month mm Bio14 Precipitation of driest month mm Bio15 Precipitation seasonality (coefficient of variation) / Bio18 Precipitation of warmest quarter mm Bio19 Precipitation of coldest quarter mm Aspect The compass direction or azimuth that a terrain surface faces / Elev Elevation, height relative to datum m Slope Angle of inclination of the slope ° Future climate data were also retrieved from WorldClim v2.1, specifically from five CMIP6 climate models (ACCESS-CM2, EC-Earth3-Veg, FIO-ESM-2-0, MPI-ESM1-2-HR, MRI-ESM2-0) that have demonstrated strong performance in East Asia [43]. The ensemble mean of these models was used to project the future distribution of S. cuneata . Future climate projections included two time periods (2041–2060, 2081–2100), under two Shared Socioeconomic Pathway (SSP) scenarios (SSP245, SSP585). Additionally, paleoclimatic datasets from WorldClim v1.4 were used to reconstruct the potential historical distribution of S. cuneata . These datasets include climate reconstructions for the LGM (about 22,000 years ago) and the MH (about 6,000 years ago), based on the ensemble mean of three climate models (CCSM4, MIROC-ESM, and MPI-ESM-P). 2.3 Species distribution models We used the biomod2 R package (https://biomodhub.github.io/biomod2/) to develop species distribution models (SDMs). A total of ten different modeling techniques were used: artificial neural network (ANN), classification tree analysis (CTA), flexible discriminant analysis (FDA), generalized additive model (GAM), generalized boosting model (GBM), generalized linear model (GLM), multiple adaptive regression splines (MARS), maximum entropy (MaxEnt), random forest (RF), and extreme gradient boosting (XGBoost). Among these, the random forest and MaxEnt models were the most frequently used. The RF algorithm aggregates the "decisions" of multiple individual trees to assign a final classification to each instance, and thereby it overcomes the limitations of a single decision tree and achieves a global optimum [44]. The MaxEnt model is a probabilistic framework based on the principle of maximum entropy, which predicts the probability distribution of species presence under given environmental conditions [11]. To optimize the MaxEnt model, two key parameters (RM—the Regularization Multiplier and FC—Feature Combination) were adjusted to control model complexity and improve accuracy. The “ENMeval” package[45] in R was used for parameter optimization, and test eight RM values (0.5 – 4, in 0.5 increments), and five feature combinations (L, LQ, LQH, LQHP, and LQHPT). Here, L, Q, H, P, and T represent linear, quadratic, hinge, product, and threshold, respectively. The optimal parameter combination was determined based on the corrected Akaike Information Criterion (AICc), and the combination with the lowest delta AICc (delta.AICc = 0) was selected to achieve the best balance between model complexity and goodness of fit. Furthermore, the model predictive performance was evaluated using the mean area under the curve (AUC) (auc.diff.avg). The final optimized settings for MaxEnt were RM = 0.5 and FC = LQHPT. For the remaining nine models, the optimal parameter settings in the Bigboss options of biomod2 were used, which should give better results than the default set. During model training, 75% of the occurrence data were randomly assigned for model training, while the remaining 25% were used for testing, with ten iterations of random validation. Model accuracy was evaluated using two key metrics: the true skill statistic (TSS) and the area under the receiver operating characteristic curve (ROC). TSS, also known as the Hanssen-Kuiper skill score, is a metric that integrates sensitivity and specificity [46]. The ROC curve plots the true positive rate against the false positive rate across different classification thresholds, and reflects the model’s classification ability at various threshold levels. The AUC value close to 1 indicates better model performance [47]. We compared the AUC and TSS values of 10 models based on the simulation results of the modern distribution. Finally, the RF model was selected for subsequent simulations of the species suitable habitat distribution. The final SDMs predicted the probability (p) of species occurrence within the study area, and this allows for the classification of suitable and unsuitable habitats. Suitable habitats were further categorized into three suitability levels—low suitable (0.2 ≤ p < 0.4), moderately suitable (0.4 ≤ p < 0.6), and extremely suitable (p ≥ 0.6) [48]. 3 Results 3.1 Current distribution We used the biomod2 package to compare the accuracy of ten species distribution models (SDMs) (Fig. 4) and their predicted species distribution results (Fig. 5). Except for the MaxEnt model, the average TSS values of the remaining nine models exceeded 0.95. Similarly, with the exception of the ANN, CTA, and MaxEnt models, the average AUC values of the other seven models were above 0.99. Although most models demonstrated high predictive accuracy, there were notable differences in the distribution of suitable habitats. We found that the RF model exhibited both high accuracy and a better alignment with the actual species distribution. Moreover, it provided a more precise identification of low suitable and moderately suitable regions, in contrast to other models that tended to produce overly binary probability classifications (Fig. 5). We simulated the global distribution probability of Sargentodoxa cuneata , but the probability of suitable habitat outside East Asia is extremely low. Therefore, only the results for the East Asian region are presented. Under current climate conditions, the total suitable habitat area is estimated at 413.76 × 10⁴ km², comprising an extremely suitable area of 177.35 × 10⁴ km², a moderately suitable area of 63.06 × 10⁴ km², and a low suitable area of 173.35 × 10⁴ km². The species suitable habitats are primarily distributed across China and Japan. The extremely suitable areas are concentrated south of the Qinling-Huaihe Line in China and in the southern and eastern regions of Japan. The moderately suitable and low suitable areas are also widely distributed in regions north of the Qinling-Huaihe Line, including Henan and Shandong provinces, as well as Yunnan Province and the Himalayan region, Hainan and Taiwan provinces, Vietnam, the Korean Peninsula, and various parts of Japan (Fig. 6). 3.2 Variable importance assessment and response curves The importance of environmental variables (Fig. 7) and the response curves of species distribution models (Fig. 8) were analyzed to identify key factors that influence the variation in suitable habitats. Among the 12 environmental variables used in the models, the most significant factors affecting the suitable habitat distribution of Sargentodoxa cuneata were precipitation of warmest quarter (bio18), precipitation of wettest month (bio13), mean diurnal range (bio2), isothermality (bio3), and annual mean temperature (bio1). Notably, precipitation-related factors (bio18, bio13) had a greater influence than temperature-related factors (bio2, bio3, bio1). The response curves of different environmental variables with respect to habitat suitability for S. cuneata revealed that eight variables showed clear optimal values for its growth: annual mean temperature (bio1), mean diurnal range (bio2), isothermality (bio3), temperature annual range (bio7), precipitation of wettest month (bio13), precipitation of driest month (bio14), precipitation of coldest quarter (bio18), and elevation. The most favorable conditions for S. cuneata growth were: annual mean temperature ≈ 15℃, mean diurnal range ≈ 7℃, isothermality ≈ 30, temperature annual range ≈ 23℃, precipitation of wettest month ≈ 400 mm, precipitation of driest month ≈ 50 mm, precipitation of coldest quarter ≈ 750 mm, and elevation ≈ 0–300 m and 1500 m. Additionally, precipitation seasonality (bio15) and slope exhibited similar trends, where higher values favored S. cuneata growth. In contrast, the suitability of S. cuneata decreased with increasing precipitation of coldest quarter (bio19). Aspect showed the least consistent effect on S. cuneata growth and had the lowest importance among the variables. Overall, S. cuneata thrives in regions with moderate temperatures, high precipitation, and pronounced precipitation seasonality, particularly at mid- and low-altitude. 3.3 Species distribution under past climatic conditions We mapped the fossil sites and found that the historical distribution range of Sargentodoxa was significantly broader than its present range. The fossil records have ages from the Middle Eocene to the Pleistocene, with occurrences in 1) the Eocene and Miocene-Pliocene of North America, 2) the Eocene, Oligocene, Miocene, and Pliocene of Europe, and 3) the Miocene and Pleistocene of Asia [34–42] (Fig. 9). The simulation of species distribution during the LGM and MH periods suggests a significant contraction in the suitable habitat compared to the present (Fig. 10). The extremely suitable area completely disappeared, while the moderately suitable area was almost entirely restricted to China, particularly to regions south of the Qinling-Huaihe Line. The suitable habitat was more contracted during the LGM period than in the MH period. During the MH period, the total suitable habitat area was 326.70 × 10⁴ km², representing a 21.04% reduction compared to the present. The moderately suitable area expanded to 104.40 × 10⁴ km², with an increase of approximately 65.55%, shifting to regions between the Yangtze-Pearl rivers, which are currently classified as extremely suitable for S. cuneata . In the LGM period, the total suitable habitat area further decreased to 283.22 × 10⁴ km², approximately 31.55% less than in the present. The moderately suitable area shrank to 56.03 × 10⁴ km², with an 11.15% decline, and was primarily confined to the border of Chongqing, Hunan, and Hubei provinces (i.e., the Wuling Mountain Range), as well as the Nanling and Wuyi mountain ranges in China. 3.4 Future suitable distribution under climate change The RF model was used to predict the potential suitable habitat of Sargentodoxa cuneata under the SSP245 and SSP585 scenarios for the periods 2041–2060 and 2081–2100 (Fig. 11b–e). The SSP245 scenario represents a moderate emission pathway, assuming global efforts to mitigate climate change, while the SSP585 scenario represents a high-emission pathway with limited climate mitigation measures. Overall, under different climate scenarios, the total suitable area is expected to increase, while the area of extremely suitable habitat is projected to decrease. Under the SSP245 scenario, the extent of change remains relatively moderate for both 2041–2060 and 2081–2100. The total suitable area is predicted to expand to 441.94 × 10⁴ km² by 2041–2060 and 451.80 × 10⁴ km² by 2081–2100 (Table 3). However, the extremely suitable area is projected to shrink to 174.90 × 10⁴ km² and 172.10 × 10⁴ km², respectively. Under the SSP585 scenario, changes are expected to be minor by 2041–2060 but become more pronounced by 2081–2100. The total suitable area will increase to 447.22 × 10⁴ km² by 2041–2060, while the extremely suitable area will decline to 173.69 × 10⁴ km². By 2081–2100, the total suitable area is predicted to expand to 525.11 × 10⁴ km², marking a 26.91% increase compared to the present. However, the extremely suitable area is expected to decline by 11.44%, shrinking to 157.06 × 10⁴ km². By calculating the difference between the future and current distributions of suitable habitat, we identified regions where habitat suitability is expected to increase or decrease under different climate scenarios (Fig. 11f–i). From the suitable distribution and changes in suitability, all scenarios suggest that suitability in the southern regions will decline, while northern areas will generally experience an increase. The suitability of coastal regions in southern China, most parts of Yunnan, and portions of the Himalayan region is expected to decrease, particularly under the SSP585 scenario by 2081–2100, where the decline is more pronounced. Conversely, regions north of the Qinling-Huaihe Line, including the North China Plain, Northeast China Plain, Hengduan Mountains, eastern Xizang Province (Fig. 2b), as well as central and northern Japan, are expected to experience an increase in habitat suitability. Notably, under the SSP585 scenario by 2081–2100, the suitability of the Northeast China Plain will increase significantly. Table 3. The different-level suitable area for the present, past, and future (10 4 km 2 ), and the areal differences were calculated compared with the present suitable area (%). Period Extremely suitable Moderate suitable Low suitable Total Area Change Area Change Area Change Area Change Modern 177.35 63.06 173.35 413.76 LGM 0 -100 56.03 -11.15 227.19 31.06 283.22 -31.55 MH 0 -100 104.40 65.55 222.31 28.24 326.70 -21.04 SSP245 2041–2060 174.90 -1.38 81.69 29.55 185.35 6.92 441.94 6.81 SSP245 2081–2100 172.10 -2.96 87.03 38.02 192.66 11.14 451.80 9.19 SSP585 2041–2060 173.69 -2.07 85.09 34.94 188.44 8.70 447.22 8.09 SSP585 2081–2100 157.06 -11.44 96.67 53.29 271.38 56.55 525.11 26.91 Using the centroid calculation function in ArcGIS, we analyzed the shifts in the centroid of the extremely suitable area under different SSP scenarios (Fig. 12). Currently, the centroid of the extremely suitable area is located in central-southern Hubei Province (112.94° E, 30.55° N). Under the SSP245 scenario, the centroid is projected to first shift southeastward to southeastern Hubei Province (114.58° E, 30.16° N), with a displacement of 163.22 km, and then move northwestward to 113.85° E, 30.29° N, covering a distance of 71.61 km. These shifts remain within Hubei Province. Under the SSP585 scenario, the centroid is expected to first move eastward to eastern Hubei Province (114.47° E, 30.59° N), with a shift of 146.55 km, and then shift further southeast beyond Hubei Province into northwestern Anhui Province, reaching coordinates of 116.31° E, 29.94° N, after an additional displacement of 190.92 km. 4 Discussion 4.1 Model performance evaluation This study compared the predictive performance of 10 species distribution models (SDMs) implemented in the biomod2 platform to simulate the current distribution of Sargentodoxa cuneata . While all models exhibited high accuracy, their predicted geographical distributions showed noticeable differences. The observed differences in predicted species distribution probabilities may be due to the complex relationships between variables. Furthermore, each model uses different methods to define the probability of species occurrence in relation to environmental variables. Based solely on AUC and TSS values, the MARS and GLM models slightly outperformed the RF model. However, in terms of geographical distribution, RF provided results that more closely matched the actual distribution pattern. The superiority of the RF model may stem from its ability to handle complex, high-dimensional data, particularly its capacity to effectively model nonlinear responses and mitigate the risk of overfitting [49]. The performance of different models varies across different species [48]. Some studies have found that RF achieves the highest accuracy [44, 50], whereas others report that MaxEnt outperforms other models [6, 51]. This highlights the inherent uncertainty across different modeling approaches and underscores the importance of selecting the most appropriate model based on the specific research objectives. Despite the strong performance of RF herein, its potential limitations should be acknowledged. Substantial biases may often arise in predictions generated by a single model under future climate scenarios. Ensemble modeling, which integrates predictions from multiple models, has been proposed as a strategy to enhance overall prediction robustness [52, 53]. However, ensemble models do not always yield the best results, as an optimized single model can sometimes outperform an ensemble approach [15]. Although RF performed well here, uncertainties in future predictions remain a concern, particularly under the complex dynamics of climate change. Therefore, future research could explore the use of ensemble modeling approaches while simultaneously optimizing the parameters of both single and ensemble models based on specific research objectives and data characteristics to achieve more robust and reliable predictions. 4.2 Potential dispersal routes of Sargentodoxa Based on available fossil records, Sargentodoxa has been distributed exclusively in the Northern Hemisphere, encompassing three major regions (Europe, North America, and Asia). The earliest fossils of Sargentodoxa were discovered along the western coast of North America from the middle Eocene, suggesting that the genus may have originated in North America (Fig. 9, Table 1) [34]. North America is also likely the diversification center for Sargentodoxa . Subsequently, Sargentodoxa spread to Europe and Asia while continuing to disperse across North America. It is reasonable to hypothesize that Sargentodoxa migrated from North America to Europe via the North Atlantic Land Bridge (NALB). There are two possible routes by which Sargentodoxa may have dispersed to Asia. The first route is a direct expansion from North America to Asia via the Bering Land Bridge (BLB). The second involves an initial spread to Europe, followed by an eastward migration across the Eurasian continent, eventually reaching Japan and southwestern China. However, the route across Eurasia remains highly uncertain due to the lack of fossil evidence. This migration could have occurred through the northern boreotropical migration via Siberia, the southern route through the Kohistan-Ladakh Island Arc (KLIA) [54], or a pathway between these two possibilities. This suggests that the dispersal of Sargentodoxa was most likely facilitated by long-distance dispersal (LDD). Both the NALB and the BLB were critical migration routes for numerous tropical, subtropical, and temperate species across the Northern Hemisphere during the Cenozoic [55, 56]. Extensive studies based on fossil records have suggested that species dispersed from North America to Europe via the NALB through Iceland [57–59] and reached Asia from North America through the BLB [60]. In the future, the discovery of more fossils of Sargentodoxa is expected to clarify its dispersal pathways and provide a more comprehensive understanding of its biogeographical history. 4.3 Identification of glacial refugia The cyclical shifts between glacial and interglacial periods since the Quaternary have profoundly influenced modern species distributions and genetic differentiation [61]. Here, species distribution models were used to simulate the suitable habitat of Sargentodoxa cuneata during the LGM and MH. The results indicate a significant southward contraction of the species suitable range compared to the present, with suitability levels decreasing. Notably, the extremely suitable area nearly disappeared entirely. During the LGM (21–18 thousand years ago, ka), global temperatures were approximately 7.0 ± 1.0°C lower than pre-industrial levels [62]. The climate in East Asia was particularly harsh, driving tropical and subtropical species southward. As S. cuneata is an indicator species of warm and humid environments, its growth was severely constrained by low temperatures, leading to a more restricted distribution and reduced suitability. Consequently, compared to the MH period, the suitable habitat during the LGM experienced further contraction, with the moderately suitable area also shrinking. Integrating fossil records from the Cenozoic era, we infer that the glacial refugia of S. cuneata were primarily located in the Nanling Mountains, the Wuyi Mountains, and the Wuling Mountains in China. These refugia overlap with those identified for East Asian relict species [19] and coincide with known refugial areas of the "living fossil" Ginkgo biloba in Southwest, East, and South China [5, 63]. Among these,the Nanling Mountains (23°37′–27°14′ N) are the largest mountain range in southern China, serving as a natural biogeographical boundary for the subtropical zone. This region is recognized as a biodiversity hotspot, providing refugium for numerous relict species due to its unique topography and ecosystem [64]. The Wuyi Mountains, situated in southeastern China along the border of Jiangxi and Fujian provinces, are characterized by highly complex topography, diverse habitats, and favorable climatic conditions. Due to these factors, the Wuyi Mountains have been widely recognized as a natural gene bank for biological species [65]. This makes their role as a refugium for S. cuneata unsurprisingly. The Wuling Mountains, a northeast-southwest trending mountain range in central China (27.28°–30.05° N, 107.02°–111.33° E), exhibit complex and diverse vegetation and serve as a biodiversity hotspot for plant species in central China [66, 67]. This region has also been identified as a refugium for many ancient plant species in China [68]. The orientation of mountain ranges is closely related to the direction of species dispersal [69]. The mountain ranges that served as glacial refugia for S. cuneata generally follow north-south, east-west, and northeast-southwest directions. These mountains facilitated its post-glacial dispersal across southern China and even East Asia. Fossil evidence further reveals that Sargentodoxa had a much wider distribution in the Paleogene (~40 Ma) and Neogene (~15 Ma), with a global presence. However, since the Quaternary (~2.58 Ma), its distribution has become restricted to China (Fig. 9, Table 1). This pattern suggests that Sargentodoxa experienced severe geographical contraction during the climatic fluctuations of the Quaternary. The existence of multiple, spatially dispersed glacial refugia likely enabled the species to persist in small populations during harsh glacial-interglacial cycles. These refugia facilitated postglacial range expansions across the Chinese subtropics, gradually shaping its current restricted distribution [27]. Given their critical role in sustaining biodiversity and ecosystem stability, these refugial regions warrant prioritized conservation efforts. 4.4 Impact of climate change on its distribution The analysis of species response curves to environmental factors sheds light on the understanding of the appropriate distribution of species under future climate change and the intrinsic reasons for the changes. Analysis of the species response curves to environmental variables revealed that precipitation plays a slightly more critical role than temperature. As a climbing plant, Sargentodoxa cuneata has high water requirements, and sufficient precipitation directly influences its growth and reproductive success. Numerous studies have identified water availability as a key limiting factor for liana growth (e.g., Jiang et al., 2011). Additionally, the response curve of precipitation seasonality (bio15) exhibits a strong linear relationship, indicating that S. cuneata thrives in environments with pronounced seasonal precipitation variation. This suggests that the species has effectively adapted to the distinct seasonal rainfall patterns associated with the East Asian monsoon climate. The ample precipitation in these regions provides the moist environment necessary for its growth. Furthermore, researchers observed a "midday depression" phenomenon in the photosynthesis of S. cuneata leaves during noon hours of summer, indirectly supporting our finding that the species suitability declines when temperatures exceed a certain threshold [71]. Interestingly, the suitable altitude range for S. cuneata may exhibit two distinct intervals: a lower altitude range (0–300 m) and a mid-altitude range (around 1500 m). Among these, the lower altitude range appears to be more suitable than the mid-altitude range, although this observation lacks sufficient literature support. Additionally, within the lower altitude range of approximately 0–500 m, habitat suitability decreases with increasing altitude. A possible explanation is that S. cuneata is more easily discovered at lower altitudes, leading to a higher number of occurrence records compared to those at mid-altitudes, which introduces bias. Future studies should incorporate a more comprehensive set of occurrence points and consider the physiological characteristics of S. cuneata to further refine the analysis of its suitable altitude range. Overall, S. cuneata is best adapted to mid-and low-altitude regions with moderate temperatures, abundant precipitation, and distinct precipitation seasonality. The response curves of environmental factors indirectly confirms that mountainous regions have been, and will continue to be, suitable habitats for S. cuneata . Mountains provide diverse topographies that offer a wide range of ecological niches, ensuring species survival across various environmental conditions. Moreover, mountain uplift influences atmospheric circulation and provides abundant orographic precipitation. The diversity of soil types and the abundance of water resources in mountainous regions further enhance the availability of essential survival resources for species. In China, the east-west- and northeast-southwest-oriented mountain ranges also block cold air from moving southward. For instance, the climate south of the Qinling-Huaihe Line is warmer and more humid. These unique geomorphological and ecological characteristics make mountains important refugia for numerous species, which play a critical role in maintaining global biodiversity during glacial and interglacial cycles [72, 73]. For example, the Hengduan Mountains in China have preserved a wide range of endemic and relict plant species by providing stable habitats [74]. This highlights the vital role that mountainous refugia play in biodiversity conservation. According to IPCC projections, by 2100, global surface temperatures are expected to rise by 2.7 °C under the moderate-emission SSP245 scenario and by 4.4 °C under the high-emission SSP585 scenario [75]. Our predictions of future habitat suitability indicate that under the SSP245 scenario, the species suitable habitat will not differ significantly from its current distribution, with a general northward shift. However, under the SSP585 scenario, the suitable range will undergo a more pronounced northward expansion. Climate warming is identified as the primary driver of S. cuneata northward expansion and southern range contraction. 5 Conclusion In this study, we compared nine machine learning models and the MaxEnt model to assess their predictive performance, ultimately selecting the RF model for simulating the past, present, and future suitable habitats of Sargentodoxa cuneata under climate change scenarios. The results indicate that S. cuneata thrives in mid- and low-altitude regions with moderate temperatures, abundant precipitation, and distinct precipitation seasonality, with its extremely suitable habitats primarily located south of the Qinling-Huaihe Line in China. Precipitation-related factors had a greater influence on distribution than temperature-related factors. During the LGM and MH periods, the species suitable range contracted significantly, with reductions of 130.54 km² and 87.06 km², respectively. The contraction was most pronounced during the LGM, when its suitable habitat was restricted to the Nanling Mountains, Wuyi Mountains, and Wuling Mountains—potential glacial refugia for the species. By 2100, model simulations indicate that under two different scenarios (SSP245 and SSP585), the total area of suitable habitat is expected to expand, with a noticeable trend of northward expansion and a corresponding contraction in the south. While the centroid of highly suitable habitats is expected to shift slightly eastward. As a representative relict plant species in China, S. cuneata shares similar responses to climate change and glacial refugia with other South Asian relict species. This study will place an important basis for future conservation strategies and climate adaptation measures for relict plants, and thus contribute to biodiversity conservation and sustainable ecosystem management. Future research combining more comprehensive fossil data, genetic analysis, and ecological modeling can further elucidate the evolutionary history, dispersal, and geographic distribution of S. cuneata . Declarations Author contributions X.L.: Conceptualization, Methodology, Investigation, Data curation, Validation, Visualization, Writing - Original Draft, Writing - Review & Editing. H.H.: Conceptualization, Methodology, Investigation, Writing - Review & Editing, Supervision, Funding acquisition. X.M.: Validation, Visualization, Writing - Review & Editing. M.L.: Validation, Visualization, Writing - Review & Editing. Z.Q.: Visualization, Writing - Review & Editing. Funding This work was supported by the Starting Grant for Introduced Talents of Sun Yat-sen University, the Fundamental Research Funds for the Central Universities, Sun Yat-sen University (No. 24qnpy021), and the General Project of Basic and Applied Basic Research of Guangzhou Bureau of Science and Technology (No. 2025A04J4384). Data availability Data is provided within the manuscript or supplementary information files. Clinical trial number Not applicable. Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Acknowledgments We thank the National Herbarium Resource Center for providing modern distribution data of Sargentodoxa cuneata . We also acknowledge Dr. Renbin Zhu for providing photos of S. cuneata . Competing Interest The authors declare no competing interests. References Fu J, Wen L. Impacts of Quaternary glaciation, geological history and geography on animal species history in continental East Asia: A phylogeographic review. Mol Ecol. 2023;32:4497–514. Guo W, Yang Y, Zhang X, Chen J, Wu S, Yang J, et al. Genomic divergence between two sister Medicago species triggered by the quaternary climatic oscillations on the Qinghai–Tibet plateau and northern China. Mol Ecol. 2023;32:3118–32. Yin Q-Y, Fan Q, Li P, Truong D, Zhao W-Y, Zhou R-C, et al. Neogene and Quaternary climate changes shaped the lineage differentiation and demographic history of Fokienia hodginsii (Cupressaceae s.l.), a Tertiary relict in East Asia. J Syst Evol. 2021;59:1081–99. Wiens JJ, Zelinka J. How many species will earth lose to climate change? Glob Chang Biol. 2024;30:e17125. Wang L, Liu J, Liu J, Wei H, Fang Y, Wang D, et al. Revealing the long-term trend of the global-scale Ginkgo biloba distribution and the impact of future climate change based on the ensemble modeling. Biodivers Conserv. 2023;32:2077–100. Kang Y, Lin F, Yin J, Han Y, Zhu M, Guo Y, et al. Projected distribution patterns of Alpinia officinarum in China under future climate scenarios: Insights from optimized Maxent and Biomod2 models. Front Plant Sci. 2025;16. Qi X, Chen C, Comes HP, Sakaguchi S, Liu Y, Tanaka N, et al. Molecular data and ecological niche modelling reveal a highly dynamic evolutionary history of the East Asian Tertiary relict Cercidiphyllum (Cercidiphyllaceae). New Phytol. 2012;196:617–30. Tang CQ, Dong Y-F, Herrando-Moraira S, Matsui T, Ohashi H, He L-Y, et al. Potential effects of climate change on geographic distribution of the Tertiary relict tree species Davidia involucrata in China. Sci Rep. 2017;7:43822. Zhao G, Cui X, Sun J, Li T, Wang Q, Ye X, et al. Analysis of the distribution pattern of Chinese Ziziphus jujuba under climate change based on optimized biomod2 and MaxEnt models. Ecological Indicators. 2021;132:108256. Elith J, Leathwick JR. Species distribution models: Ecological explanation and prediction across space and time. Annu Rev Ecol Evol Syst. 2009;40 Volume 40, 2009:677–97. Phillips SJ, Anderson RP, Schapire RE. Maximum entropy modeling of species geographic distributions. Ecol Modell. 2006;190:231–59. Breiman L. Random forests. Mach Learn. 2001;45:5–32. Nelder JA, Wedderburn RWM. Generalized linear models. J R Stat Soc Ser A. 1972;135:370–84. Hastie T, Tibshirani R. Generalized additive models: Some applications. J Am Stat Assoc. 1987;82:371–86. Hao T, Elith J, Lahoz-Monfort JJ, Guillera-Arroita G. Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models. Ecography. 2020;43:549–58. Li X, Wang Y. Applying various algorithms for species distribution modelling. Integr Zool. 2013;8:124–35. Médail F, Diadema K. Glacial refugia influence plant diversity patterns in the Mediterranean Basin. J Biogeogr. 2009;36:1333–45. Hampe A, Rodríguez‐Sánchez F, Dobrowski S, Hu FS, Gavin DG. Climate refugia: From the Last Glacial Maximum to the twenty‐first century. New Phytol. 2013;197:16–8. Tang CQ, Matsui T, Ohashi H, Dong Y-F, Momohara A, Herrando-Moraira S, et al. Identifying long-term stable refugia for relict plant species in East Asia. Nat Commun. 2018;9:4488. Petit RJ, Aguinagalde I, de Beaulieu J-L, Bittkau C, Brewer S, Cheddadi R, et al. Glacial refugia: Hotspots but not melting pots of genetic diversity. Science. 2003;300:1563–5. Singh PB, Mainali K, Jiang Z, Thapa A, Subedi N, Awan MN, et al. Projected distribution and climate refugia of endangered Kashmir musk deer Moschus cupreus in greater Himalaya, South Asia. Sci Rep. 2020;10:1511. Chen DZ, Tatemi S. Lardizabalaceae. Flora of China. 2001;6:440–54. Zhao X, Ke H, Yu H. Studies on pharmacological effect and clinical application of Sargentodoxa cuneata . Guiding Journal of Traditional Chinese Medicine and Pharmacy. 2014;20:41–3. Zhang W, Sun C, Zhou S, Zhao W, Wang L, Sheng L, et al. Recent advances in chemistry and bioactivity of Sargentodoxa cuneata . J Ethnopharmacol. 2021;270:113840. Wang Y, Zhang B, Liu S, Xu E, Wang Z. The traditional herb Sargentodoxa cuneata alleviates DSS-induced colitis by attenuating epithelial barrier damage via blocking necroptotic signaling. J Ethnopharmacol. 2024;319:117373. Xu F, Yu P, Wu H, Liu M, Liu H, Zeng Q, et al. Aqueous extract of Sargentodoxa cuneata alleviates ulcerative colitis and its associated liver injuries in mice through the modulation of intestinal flora and related metabolites. Front Microbiol. 2024;15. Tian S, Lei S-Q, Hu W, Deng L-L, Li B, Meng Q-L, et al. Repeated range expansions and inter-/postglacial recolonization routes of Sargentodoxa cuneata (oliv.) rehd. et wils. (Lardizabalaceae) in subtropical China revealed by chloroplast phylogeography. Mol Phylogenet Evol. 2015;85:238–46. Manchester SR, Chen Z-D, Lu A-M, Uemura K. Eastern Asian endemic seed plant genera and their paleogeographic history throughout the Northern Hemisphere. Journal of Systematics and Evolution. 2009;47:1–42. Zhou Z, Arata M. Fossil history of some endemic seed plants of east asiaand its phytogeographical significance. Acta Bot Yunnan. 2005;27:449–70. GBIF.org. GBIF occurrence downloaded on 23rd december 2024. Available at https://doi.org/1015468/dl.zsphbx. 2024. Zizka A, Silvestro D, Andermann T, Azevedo J, Duarte Ritter C, Edler D, et al. CoordinateCleaner: Standardized cleaning of occurrence records from biological collection databases. Methods Ecol Evol. 2019;10:744–51. Warren DL, Matzke NJ, Cardillo M, Baumgartner JB, Beaumont LJ, Turelli M, et al. ENMTools 1.0: An R package for comparative ecological biogeography. Ecography. 2021;44:504–11. Thuiller W, Georges D, Engler R, Breiner F, Georges MD, Thuiller CW. Package ‘biomod2.’ Species distribution modeling within an ensemble forecasting framework. 2016;10:1600-0587.2008. Manchester SR. Biogeographical relationships of north American Tertiary floras. Ann Mo Bot Gard. 1999;86:472–522. Song Z. Late Cenozoic palyno-flora from Zhaotong, Yunnan. Proceedings of the Nanjing Institute of Geology and Palaeontology, Chinese Academy of Sciences. 1988;0:1–108. Mai HD. Die mittelmiozänen und obermiozänen floren aus der meuroer und raunoer folge in der lausitz. Teil II: Dicotyledonen. Palaeontographica Abteilung B. 2001;257:35–174. Tiffney BH. Fruits and seeds of the Tertiary Brandon Lignite. Vii. Sargentodoxa (Sargentodoxaceae). Am J Bot. 1993;80:517–23. McNair D, Stults D, Axsmith B, Alford M, Starnes J. Preliminary investigation of a diverse megafossil floral assemblage from the middle Miocene of southern Mississippi, USA. Palaeontol Electronica. 2019;22.2.40A:1–30. Momohara A. Change of paleovegetation caused by topographic change in and around a sedimentary basin of the Upper Miocene Tokiguchi Porcelain Clay Formation, central Japan. Geoscience Report of the Shimane University. 2001;20:49. Mead JI, Schubert BW, Wallace SC, Swift SL. Helodermatid lizard from the Mio-Pliocene oak-hickory forest of Tennessee, eastern USA, and a review of monstersaurian osteoderms. Acta Palaeontol Polonica. 2012;57:111–21. Geissert F, Gregor HJ, Mai DH. Die “saugbaggerflora”: Eine frucht- und samenflora aus dem grenzbereich miozän-pliozän von sessenheim im elsass (frankreich). Forschungen aus den Naturwissenschaften; 1990. Martinetto E. Studies on some exotic elements of the Pliocene floras of Italy. Palaeontographica Abteilung B. 2001;:149–66. Zhang M-Z, Xu Z, Han Y, Guo W. Evaluation of CMIP6 models toward dynamical downscaling over 14 CORDEX domains. Clim Dyn. 2024;62:4475–89. Zhao Z, Xiao N, Shen M, Li J. Comparison between optimized MaxEnt and random forest modeling in predicting potential distribution: A case study with Quasipaa boulengeri in China. Sci Total Environ. 2022;842:156867. Kass JM, Muscarella R, Galante PJ, Bohl CL, Pinilla‐Buitrago GE, Boria RA, et al. ENMeval 2.0: Redesigned for customizable and reproducible modeling of species’ niches and distributions. Methods Ecol Evol. 2021;12:1602–8. Allouche O, Tsoar A, Kadmon R. Assessing the accuracy of species distribution models: Prevalence, kappa and the true skill statistic (TSS). J Appl Ecol. 2006;43:1223–32. Shabani F, Kumar L, Ahmadi M. Assessing accuracy methods of species distribution models: AUC, specificity, sensitivity and the true skill statistic. Global Journal of Human-Social Science: B Geography, Geo-Sciences, Environmental Science & Disaster Management. 2018;18. Wang P, Luo W, Zhang Q, Han S, Jin Z, Liu J, et al. Assessing the impact of climate change on three populus species in China: Distribution patterns and implications. Glob Ecol Conserv. 2024;50:e02853. Valavi R, Elith J, Lahoz‐Monfort JJ, Guillera‐Arroita G. Modelling species presence‐only data with random forests. Ecography. 2021;44:1731–42. Li C, Luo G, Yue C, Zhang L, Duan Y, Liu Y, et al. Distribution patterns and potential suitable habitat prediction of Ceracris kiangsu (Orthoptera: Arcypteridae) under climate change - a case study of China and Southeast Asia. Sci Rep. 2024;14:20580. Cao G, Yuan X, Shu Q, Gao Y, Wu T, Xiao C, et al. Prediction of the potentially suitable areas of Eucommia ulmoides Oliver in China under climate change based on optimized Biomod2 and MaxEnt models. Front Plant Sci. 2024;15. Wang D, Shi C, Alamgir K, Kwon S, Pan L, Zhu Y, et al. Global assessment of the distribution and conservation status of a key medicinal plant ( Artemisia annua L.): The roles of climate and anthropogenic activities. Sci Total Environ. 2022;821:153378. Wu Y, Shen J, Deane DC, Yu H, Yu F, Wang X, et al. Future extreme climate events threaten alpine and subalpine woody plants in China. Earth’s Future. 2025;13:e2024EF005147. Gao Y, Song A, Cai W-J, Spicer RA, Zhang R, Liu J, et al. Tibetan Plateau palm fossils prove the Kohistan-Iadakh Island Arc is a floristic steppingstone between Gondwana and Laurasia. Rev Palaeobot Palynol. 2025;334:105255. Hopkins DM. Cenozoic history of the Bering Land Bridge. Science. 1959;129:1519–28. Tiffney BH. The Eocene North Atlantic Land Bridge: Its importance in Tertiary and modern phytogeography of the Northern Hemisphere. J Arnold Arbor. 1985;66:243–73. Denk T, Grímsson F, Zetter R, Símonarson LA. The biogeographic history of Iceland – the North Atlantic Land Bridge revisited. In: Denk T, Grimsson F, Zetter R, Símonarson LA, editors. Late Cainozoic Floras of Iceland: 15 Million Years of Vegetation and Climate History in the Northern North Atlantic. Dordrecht: Springer Netherlands; 2011. p. 647–68. Jia L-B, Manchester SR, Su T, Xing Y-W, Chen W-Y, Huang Y-J, et al. First occurrence of Cedrelospermum (Ulmaceae) in Asia and its biogeographic implications. J Plant Res. 2015;128:747–61. Jiang Y, Gao M, Meng Y, Wen J, Ge X-J, Nie Z-L. The importance of the North Atlantic land bridges and eastern Asia in the post-Boreotropical biogeography of the Northern Hemisphere as revealed from the poison ivy genus ( Toxicodendron , Anacardiaceae). Mol Phylogenet Evol. 2019;139:106561. Wen J, Nie Z, Ickert‐Bond SM. Intercontinental disjunctions between eastern Asia and western North America in vascular plants highlight the biogeographic importance of the Bering land bridge from late Cretaceous to Neogene. J Syst Evol. 2016;54:469–90. Hewitt G. The genetic legacy of the Quaternary ice ages. Nature. 2000;405:907–13. Osman MB, Tierney JE, Zhu J, Tardif R, Hakim GJ, King J, et al. Globally resolved surface temperatures since the Last Glacial Maximum. Nature. 2021;599:239–44. Zhao Y-P, Fan G, Yin P-P, Sun S, Li N, Hong X, et al. Resequencing 545 ginkgo genomes across the world reveals the evolutionary history of the living fossil. Nat Commun. 2019;10:4201. Tian S, Kou Y, Zhang Z, Yuan L, Li D, López-Pujol J, et al. Phylogeography of Eomecon chionantha in subtropical China: The dual roles of the Nanling Mountains as a glacial refugium and a dispersal corridor. BMC Evol Biol. 2018;18:20. Chen L, Cao X, Pan T, Lei P, Zeng L, Li M, et al. Conservation status and prioritization of rare and endangered plants in Jiangxi Wuyishan National Nature Reserve. Journal of Nanjing Forestry University (Natural Sciences Edition). 2024;48:39. Sun Z, Yang L, Kong H, Kang M, Wang J. Phylogeographical patterns match the floristic subdivisions: The diversification history of a widespread herb in subtropical China. Ann Bot. 2024;134:1263–76. Wang C, Zhou T, Qin Y, Zhou G, Fei Y, Xu Y, et al. Wuling mountains function as a corridor for woody plant species exchange between northern and southern central China. Front Ecol Evol. 2022;10. Qi C, Yu X, Zheng Z, Yin G. Notes on endemic seed plants in the floristic region of central China. Journal of Central-South Forestry College. 1998;18:1–4. Xiao Y, Li X-J, Jiang X-L, Li C, Li X-P, Li W-P, et al. Spatial genetic patterns and distribution dynamics of Begonia grandis (Begoniaceae), a widespread herbaceous species in China. Front Plant Sci. 2023;14. Jiang H, Zhou G-Y, Huang Y-H, Liu S-Z, Tang X-L. Photosynthetic characteristics of canopy-dwelling vines in lower subtropical evergreen broad-leaved forest and response to environmental factors: Photosynthetic characteristics of canopy-dwelling vines in lower subtropical evergreen broad-leaved forest and response to environmental factors. Chin J Plant Ecol. 2011;35:567–76. Jin Z, Ke S, Zhong Z. Studies on characteristics of leaves growth and photosynthesis physioecology of Sargentodoxa cuneata . Bulletin of Botanical Research. 2002;22:184–9. Antonelli A, Kissling WD, Flantua SGA, Bermúdez MA, Mulch A, Muellner-Riehl AN, et al. Geological and climatic influences on mountain biodiversity. Nature Geosci. 2018;11:718–25. Hoorn C, Mosbrugger V, Mulch A, Antonelli A. Biodiversity from mountain building. Nature Geosci. 2013;6:154–154. Sun H, Zhang J, Deng T, Boufford DE. Origins and evolution of plant diversity in the Hengduan Mountains, China. Plant Divers. 2017;39:161–6. Intergovernmental Panel on Climate Change (IPCC). Climate change 2021 – the physical science basis: Working group I contribution to the sixth assessment report of the intergovernmental panel on climate change. 1st edition. Cambridge University Press; 2021. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6516979","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":447225147,"identity":"ea14f28d-88bc-43b0-9e9e-0e0109d8128e","order_by":0,"name":"Xuanqi Liu","email":"","orcid":"","institution":"School of Geography and Planning, Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Xuanqi","middleName":"","lastName":"Liu","suffix":""},{"id":447225150,"identity":"ee2fd871-3d1f-47f7-9672-154833aa6cfc","order_by":1,"name":"Huasheng Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYBACAxDB2AAimQ8ACQkZUrSwJYC08JCihQfMJqzFXCLHTPLnjsPy5vxrPr+6UWPBw8B++OgGfFosZ6SlSfOeOWy4c8bbbdY5x4AO40lLu4HXYTeSj0kzth1m3HDj7DbjHDagFgkeMwJaEtskf7Ydtt9w48wz45x/RGlJPibB23Y4ccP5HubHuW1EaLHseZZszduWnrzhBpsZc26fBA8bIb+Ys+cY3vzZZm274fzhx59zvtXJ8bMfPoZXCwJIJLBJgGg24pSDAP8B5g/Eqx4Fo2AUjIKRBAAwJks/IVS2twAAAABJRU5ErkJggg==","orcid":"","institution":"School of Geography and Planning, Sun Yat-sen University","correspondingAuthor":true,"prefix":"","firstName":"Huasheng","middleName":"","lastName":"Huang","suffix":""},{"id":447225151,"identity":"c62d5e63-667e-4368-9a5e-252d8b001ec5","order_by":2,"name":"Xia Meng","email":"","orcid":"","institution":"School of Geography and Planning, Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Xia","middleName":"","lastName":"Meng","suffix":""},{"id":447225152,"identity":"751d3cfd-458a-43a2-9a6f-ee61b72a6cd9","order_by":3,"name":"Minqiao Li","email":"","orcid":"","institution":"School of Geography and Planning, Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Minqiao","middleName":"","lastName":"Li","suffix":""},{"id":447225153,"identity":"d7548015-e852-478a-bb81-3807b33d3158","order_by":4,"name":"Zeyu Qin","email":"","orcid":"","institution":"School of Geography and Planning, Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Zeyu","middleName":"","lastName":"Qin","suffix":""}],"badges":[],"createdAt":"2025-04-24 04:38:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6516979/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6516979/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":81688567,"identity":"8de28641-3441-4611-ad64-c1da3c0546c7","added_by":"auto","created_at":"2025-04-30 11:08:56","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":31007892,"visible":true,"origin":"","legend":"\u003cp\u003eLeaves (a), fruits (b), rattan (c), and medicinal slices (d) of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e. The medicinal slices are made from the cleaned and dried stems. Photos (a)–(c) courtesy of Dr. Renbin Zhu, (d) taken by Dr. Huasheng Huang.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/325a47085534875d03967fc6.png"},{"id":81688556,"identity":"47bfec63-8e06-4650-be72-6219ca568edd","added_by":"auto","created_at":"2025-04-30 11:08:55","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":889401,"visible":true,"origin":"","legend":"\u003cp\u003e(a) Modern cleaned occurrences of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e obtained from GBIF (http://www.gbif.org) and CVH (http://www.cvh.ac.cn).(b) The important geographical units in China mentioned in this study include mountain ranges (indicated by red dashed circles), the Yangtze and Pearl rivers (blue lines), and the Qinling-Huaihe Line (orange dashed line).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/d24f4f70f93a88e4b2abd08f.png"},{"id":81689557,"identity":"db859c22-f73d-43d1-ade9-3f7a80aeccc4","added_by":"auto","created_at":"2025-04-30 11:24:56","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":14975602,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation of 19 climate variables. Red and blue colors indicate positive and negative correlations, respectively. The intensity of the color and the size of the corresponding circle increase with the correlation coefficient. For example, a correlation coefficient of 0.99 between bio13 and bio16, suggests that only one of them should be retained in subsequent analyses. Abbreviations: bio1 = annual mean temperature, bio2 = mean diurnal range, bio3 = isothermality, bio4 = temperature seasonality, bio5 = max temperature of warmest month, bio6 = min temperature of coldest month, bio7 = temperature annual range, bio8 = mean temperature of wettest quarter, bio9 = mean temperature of driest quarter, bio 10 = mean temperature of warmest quarter, bio11 = mean temperature of coldest quarter, bio12 = annual precipitation, bio13 = precipitation of wettest month, bio14 = precipitation of driest month, bio15 = precipitation seasonality, bio16 = precipitation of wettest quarter, bio17 = precipitation of driest quarter, bio18 = precipitation of warmest quarter, bio19 = precipitation of coldest quarter.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/8acd62c8c3a4c78d35c207ac.png"},{"id":81689556,"identity":"ace65afc-35b4-4742-bffb-fce6bc6d2340","added_by":"auto","created_at":"2025-04-30 11:24:55","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1440009,"visible":true,"origin":"","legend":"\u003cp\u003eAccuracy of the ten SDM models, and their TSS and AUC values. The error bars extending from each point represent the range of variability.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/48861f29e9bf85743b3da6ce.png"},{"id":81688558,"identity":"6cb3f757-0a91-4999-8ecc-46fdb293e9fe","added_by":"auto","created_at":"2025-04-30 11:08:55","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1093305,"visible":true,"origin":"","legend":"\u003cp\u003eModern distribution of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e in the ten SDMs.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/c0833e39b57759b48ce4ee05.png"},{"id":81688818,"identity":"700476fa-ad17-4990-b108-bfcc431e963a","added_by":"auto","created_at":"2025-04-30 11:16:55","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":448035,"visible":true,"origin":"","legend":"\u003cp\u003eModern distribution of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e using a random forest model. Abbreviations—province in China: CQ = Chongqing, GZ = Guizhou, GX = Guangxi, HeN = Henan, HB = Hubei, HuN = Hunan, GD = Guangdong, HN = Hainan, SD = Shandong, JS = Jiangsu, AH = Anhui, JX = Jiangxi, ZJ = Zhejiang, FJ = Fujian, TW = Taiwan; Country: N. Korea = North Korea, i.e., the Democratic People's Republic of Korea; S. Korea = South Korea, namely the Republic of Korea.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/327f209d7b4900b64d0dfcbc.png"},{"id":81688559,"identity":"ad9bf3b8-2b03-4c8f-a4e3-908122c6984d","added_by":"auto","created_at":"2025-04-30 11:08:55","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":896854,"visible":true,"origin":"","legend":"\u003cp\u003eThe importance of different environmental variables. The full names and units of each variable are shown in Table 2. The gray short lines represent the fluctuation range from multiple random simulations.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/aa3450be534680e8b37b5624.png"},{"id":81688563,"identity":"fade318b-5564-4caf-8508-634411dff242","added_by":"auto","created_at":"2025-04-30 11:08:56","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":5398392,"visible":true,"origin":"","legend":"\u003cp\u003eEnvironmental variable response curve. The full names and units of each variable are shown in Table 2. The light red background indicates the fluctuation range of multiple random simulations.\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/f91ccb9228923e6846143d55.png"},{"id":81688564,"identity":"a0518015-0626-4c32-8253-bbc8c6c181ad","added_by":"auto","created_at":"2025-04-30 11:08:56","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":16839911,"visible":true,"origin":"","legend":"\u003cp\u003eStratigraphical ranges and global distribution, and proposed dispersal routes of \u003cem\u003eSargentodoxa\u003c/em\u003e fossils. (a) Stratigraphical ranges of \u003cem\u003eSargentodoxa\u003c/em\u003e fossils records. All fossil records are within the Cenozoic. Abbreviations: Q = Quaternary, N = Neogene, P = Paleogene, Pleist. – Holo. = Pleistocene–Holocene period. The short vertical lines following the symbols indicate the age uncertainty. The Miocene fossil record (no. 6) in Japan has an age of 10 Ma. Only the latest (Pliocene–Pleistocene) record (no. 10) in China is a microfossil (pollen) record, which also has an uncertain age. (b) Global distribution of \u003cem\u003eSargentodoxa\u003c/em\u003e fossils across the Northern Hemisphere and possible dispersal routes. The routes with question marks indicate considerable uncertainty. The information of all fossil records is provided in Table 1. The numerical labels in the figure correspond to the Record ID in Table 1.\u003c/p\u003e","description":"","filename":"image9.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/b4af2df9810c240b416b3b60.png"},{"id":81688820,"identity":"137bc163-e769-4d60-aabf-82c6dabad252","added_by":"auto","created_at":"2025-04-30 11:16:55","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":459053,"visible":true,"origin":"","legend":"\u003cp\u003eSuitable areas of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e in the past suggested by random forest model. (a) Middle Holocene. (b) Last Glacial Maximum. The red circles showing the localities of Wuling Mountains, Wuyi Mountains, and Nanling Mountains, which played as important refugia for \u003cem\u003eS. cuneata\u003c/em\u003e during the Last Glacial Maximum period. The blue lines are Yangtze and Pearl rivers.\u003c/p\u003e","description":"","filename":"image10.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/4f5bd9abbf9a607d2ccbfaf6.png"},{"id":81688823,"identity":"033d307b-7c90-41dc-8ade-34c50f72ad43","added_by":"auto","created_at":"2025-04-30 11:16:56","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":791333,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution and changes of future suitable areas compared to its current distribution for \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e under different scenarios. (a) Modern suitable habitat areas; (b), (c), (d), and (e) show future suitable areas; (f), (g), (h), and (i) illustrate future habitat suitability changes relative to the current distribution.\u003c/p\u003e","description":"","filename":"image11.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/134d53d470962a2835255cf6.png"},{"id":81688822,"identity":"178bef67-2171-442e-b811-8420cd081375","added_by":"auto","created_at":"2025-04-30 11:16:56","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":407422,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution and shifts in the centroid of the extremely suitable area for \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e under future climate scenarios. The green line represents movement in the path of SSP245, the red line is SSP585, and the arrow suggests the direction of movement.\u003c/p\u003e","description":"","filename":"image12.png","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/88782fece83a7cd24a306a7f.png"},{"id":82129119,"identity":"afca4911-5f46-4fad-ae03-d8c6fcf13726","added_by":"auto","created_at":"2025-05-07 05:02:01","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":63012850,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6516979/v1/98651fcb-99bf-42b2-932d-f180e0a950d5.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"The past, present and future distribution of Sargentodoxa Rehder \u0026 E.H.Wilson: Perspectives from fossil record and species distribution models","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eOver the past few decades, the global average temperature has steadily risen, accompanied by an increasing frequency of extreme climatic events. This have significantly impacted ecosystem stability and species adaptive capacities. Since the Quaternary period, glacial-interglacial cycles have driven habitat shifts in many extant species across the Northern Hemisphere [1], and also led to genomic divergence [2, 3]. Climate change not only reshapes the habitat suitability but also contributes to habitat loss, population declines, and even extinction [4]. In the face of escalating climate change, species geographical distribution patterns are undergoing profound transformations. Accurately predicting these distributional shifts and glacial refugia is essential for biodiversity conservation, biological resource management, and sustainable development [5].\u003c/p\u003e\n\u003cp\u003eThe species distribution models (SDMs), also known as ecological (or environmental) niche models (ENMs), have become a fundamental tool for assessing the influence of environmental factors on species distributions. The SDMs have been widely applied in tracing ecological indication and biogeographical history [6\u0026ndash;9]. They estimate potential suitable habitats for species across different temporal scales by modeling the relationship between known species occurrences and environmental variables [10]. The SDM algorithms that have been commonly used include maximum entropy model (MaxEnt) [11], random forest (RF) [12], generalized linear models (GLM) [13], and generalized additive models (GAM) [14], among others. The selection of the most appropriate modeling approach depends on the characteristics and objectives of the study [15, 16].\u003c/p\u003e\n\u003cp\u003eGlacial refugia refer to regions that provided climatically suitable conditions for species persistence during global glacial periods, such as the Last Glacial Maximum (LGM) [17]. These areas are sometimes also referred to as climate refugia [18]. Plants, highly sensitive to climatic fluctuations, rely on refugia not only for survival but also as centers of genetic differentiation and adaptive evolution. Consequently, glacial refugia play a critical role in shaping evolutionary processes, maintaining biodiversity, and influencing present-day species distribution patterns [19]. During glacial periods, many species experienced severe range contractions or even local extinctions due to temperature declines and ice sheet expansion. However, certain regions with complex topography and diverse microclimates, such as Southwest China [19], the Mediterranean Basin [20] and the Himalayas [21], served as crucial refugia for subtropical and temperate plants. These areas not only facilitated species survival during glacial episodes but also acted as postglacial centers of recolonization, shaping contemporary distribution patterns. By using species distribution models (SDMs), we can delineate the current suitable habitat range of a species, reconstruct its historical distribution patterns under past climatic conditions, and thus identify its glacial refugia. Additionally, the SDMs can be used to predict potential habitat shifts under future climate change scenarios, and this provides evidences for the development of effective conservation strategies.\u003c/p\u003e\n\u003cp\u003eThe genus \u003cem\u003eSargentodoxa\u0026nbsp;\u003c/em\u003eRehder \u0026amp; E.H.Wilson belongs to the family Lardizabalaceae, and is a monotypic genus containing only \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e (Oliv.) Rehd. et Wils (Fig. 1). \u003cem\u003eS. cuneata\u003c/em\u003e is a relict species endemic to China, and primarily distributed in subtropical regions, with occasional occurrences in northern Laos and Vietnam. It commonly grows in well-lit open forests, forest edges, and shrublands on mountain slopes at elevations of 100\u0026ndash;1000 m [22]. \u003cem\u003eS. cuneata\u003c/em\u003e holds significant medicinal value [23]. Its bioactive compounds exhibit anti-inflammatory, antibacterial, and antitumor properties, which are helpful for treating various diseases, particularly arthritis and sepsis [24]. Consequently, the sustainable utilization and conservation of this species have attracted considerable attention. Currently, research on medicinal properties of \u003cem\u003eS. cuneata\u003c/em\u003e is relatively abundant [25, 26], while studies on its spatiotemporal distribution remain limited [27].\u003c/p\u003e\n\u003cp\u003eClimate change is expected to have profound impacts on the spatiotemporal distribution of \u003cem\u003eS. cuneata\u003c/em\u003e. Fossil evidence indicates that \u003cem\u003eS. cuneata\u003c/em\u003e has once broader distribution (across the globe) over the past several million years than at present. However, following large-scale extinctions in the Americas and Europe, its habitat is now restricted to certain regions of East Asia [28, 29]. The identification of historical glacial refugia for S. cuneata can shed light on why its current distribution is limited to China and offer insights into its potential adaptability to different climatic conditions. Furthermore, these refugial regions may still serve as contemporary genetic diversity hotspots for \u003cem\u003eS. cuneata\u003c/em\u003e. Recognizing and protecting these areas is crucial for preserving genetic resources and enhancing the species resilience to future climate change.\u003c/p\u003e\n\u003cp\u003eAlthough \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e is a relict species of significant ecological and medicinal value, studies on its suitable habitat distribution remain scarce. Therefore, this study aims to address the following key questions: How does the suitable habitat of \u003cem\u003eS. cuneata\u003c/em\u003e change from the past to present and the future in response to climate change? Where were the glacial refugia for \u003cem\u003eS. cuneata\u003c/em\u003e during the ice age? To answer these questions, we apply species distribution models (SDMs) to simulate the suitable habitat distribution of \u003cem\u003eS. cuneata\u003c/em\u003e under past, present, and future climatic scenarios. We analyze changes in suitable habitat and the species responses to climate change. We also integrate fossil evidence to untangle its biogeographical history. We hope this study will provide valuable insights into the biogeographical history of \u003cem\u003eS. cuneata\u003c/em\u003e, and shed light on maintaining the stability of East Asian ecosystems and promoting biodiversity conservation under the scenario of global climate change.\u003c/p\u003e"},{"header":"2 Materials and Methods","content":"\u003ch2\u003e2.1 Data collection\u003c/h2\u003e\n\u003cp\u003eThe modern distribution data of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e were obtained from the Global Biodiversity Information Facility (http://www.gbif.org, GBIF.org, 2024) and the National Plant Specimen Resource Center (CVH, http://www.cvh.ac.cn/). To ensure data quality, outliers and duplicate records were removed using CoordinateCleaner [31] and ENMTools packages [32] in R. This ensures that a maximum of one occurrence point was retained per 5-minute grid cell. Additionally, two anomalous occurrence points from Sri Lanka and northeastern China were identified. After verification with Flora Reipublicae Popularis Sinicae (FRPS, https://www.iplant.cn/) and Plants of the World Online (POWO, https://powo.science.kew.org/), these two records were deemed erroneous and subsequently removed. This resulted in a total of 794 geographically valid occurrence records for ecological niche modeling (Fig. 2).\u003c/p\u003e\n\u003cp\u003eTo facilitate robust species distribution modeling, 2,400 pseudo-absence points were generated. Using the disk method in the biomod2 package [33], pseudo-absence points were selected within a circular buffer around presence points, with a minimum distance of 10 km.\u003c/p\u003e\n\u003cp\u003eFossil records of the relict species \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e play a crucial role in identifying its potential past refugia. A total of 10 fossil records were retrieved from the literature, including nine macrofossil records (e.g., Manchester, 1999) primarily consisting of seeds, and one Chinese pollen record [35]. These fossil data were used to validate and disentangle the historical distribution patterns of \u003cem\u003eS. cuneata\u003c/em\u003e (Table 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1.\u003c/strong\u003e Fossil records of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecord ID\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eOrgan type\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eForm-species\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eLatitude\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eLongitude\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCountry\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eReference\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSeed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eS. globosa\u0026nbsp;\u003c/em\u003eManchester\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e44.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e-120.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUSA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMiddle Eocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eManchester, 1999\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSeed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eS. lusatica\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e51.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e12.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eGermany\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eLate Eocene\u0026ndash;Late Oligocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMai, 2001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eFruit/seed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eS. cuneata\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e43.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e-73.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUSA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eEarly Miocene (mid-Tertiary)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eTiffney, 1993\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSeed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eS.\u0026nbsp;\u003c/em\u003esp.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e31.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e-89.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUSA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMiddle Miocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMcNair et al., 2019\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSeed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eS. lusatica\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e51.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e14.20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eGermany\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMiddle Miocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMai, 2001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMacrofossil\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eS. cuneata\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e35.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e137.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eJapan\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10 Ma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMomohara, 2001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMacrofossil\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eSargentodoxa\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e36.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e-82.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUSA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eLate Miocene\u0026ndash;Early Pliocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMead et al., 2012\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSeed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eSargentodoxa\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e48.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e7.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eFrench\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eLate Miocene\u0026ndash;Early Pliocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eGeissert et al., 1990\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSeed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eSargentodoxa\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e42.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e13.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eItaly\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePliocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMartinetto, 2001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePollen\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eS. cuneata\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e27.34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e103.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eChina\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePliocene\u0026ndash;Pleistocene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSong, 1988\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e2.2 Environmental variables\u003c/h2\u003e\n\u003cp\u003eWe selected 19 bioclimatic variables that could potentially influence the distribution of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e from the WorldClim Database version 2.1 (www.worldclim.org), with a spatial resolution of 5 minutes. To prevent model overfitting due to multicollinearity among environmental variables, we calculated the correlation coefficients between variables (Fig. 3). Variables showing a correlation coefficient with an absolute value less than 0.8 (|r| \u0026lt; 0.8) were retained, and this results in the selection of nine bioclimatic variables. Additionally, elevation data were sourced from WorldClim, and topographic factors including slope and aspect were calculated using ArcGIS. This leads to a final set of 12 environmental variables for modeling (Table 2).\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2.\u003c/strong\u003e Environmental factors for modeling SDMs.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVariables\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnits\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003eAnnual mean Temperature\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e℃\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003eMean diurnal range (mean of monthly (max temp-min temp))\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e℃\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003eIsothermality (bio2/bio7) (\u0026times;100)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e/\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003eTemperature annual range (bio5-bio6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e℃\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003ePrecipitation of wettest month\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003emm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003ePrecipitation of driest month\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003emm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003ePrecipitation seasonality (coefficient of variation)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e/\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003ePrecipitation of warmest quarter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003emm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBio19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003ePrecipitation of coldest quarter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003emm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eAspect\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003eThe compass direction or azimuth that a terrain surface faces\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e/\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eElev\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003eElevation, height relative to datum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003em\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eSlope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 305px;\"\u003e\n \u003cp\u003eAngle of inclination of the slope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026deg;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eFuture climate data were also retrieved from WorldClim v2.1, specifically from five CMIP6 climate models (ACCESS-CM2, EC-Earth3-Veg, FIO-ESM-2-0, MPI-ESM1-2-HR, MRI-ESM2-0) that have demonstrated strong performance in East Asia [43]. The ensemble mean of these models was used to project the future distribution of \u003cem\u003eS. cuneata\u003c/em\u003e. Future climate projections included two time periods (2041\u0026ndash;2060, 2081\u0026ndash;2100), under two Shared Socioeconomic Pathway (SSP) scenarios (SSP245, SSP585). Additionally, paleoclimatic datasets from WorldClim v1.4 were used to reconstruct the potential historical distribution of \u003cem\u003eS. cuneata\u003c/em\u003e. These datasets include climate reconstructions for the LGM (about 22,000 years ago) and the MH (about 6,000 years ago), based on the ensemble mean of three climate models (CCSM4, MIROC-ESM, and MPI-ESM-P).\u003c/p\u003e\n\u003ch2\u003e2.3 Species distribution models\u003c/h2\u003e\n\u003cp\u003eWe used the biomod2 R package (https://biomodhub.github.io/biomod2/) to develop species distribution models (SDMs). A total of ten different modeling techniques were used: artificial neural network (ANN), classification tree analysis (CTA), flexible discriminant analysis (FDA), generalized additive model (GAM), generalized boosting model (GBM), generalized linear model (GLM), multiple adaptive regression splines (MARS), maximum entropy (MaxEnt), random forest (RF), and extreme gradient boosting (XGBoost). Among these, the random forest and MaxEnt models were the most frequently used. The RF algorithm aggregates the \u0026quot;decisions\u0026quot; of multiple individual trees to assign a final classification to each instance, and thereby it overcomes the limitations of a single decision tree and achieves a global optimum [44]. The MaxEnt model is a probabilistic framework based on the principle of maximum entropy, which predicts the probability distribution of species presence under given environmental conditions [11]. To optimize the MaxEnt model, two key parameters (RM\u0026mdash;the Regularization Multiplier and FC\u0026mdash;Feature Combination) were adjusted to control model complexity and improve accuracy. The \u0026ldquo;ENMeval\u0026rdquo; package[45] in R was used for parameter optimization, and test eight RM values (0.5 \u0026ndash; 4, in 0.5 increments), and five feature combinations (L, LQ, LQH, LQHP, and LQHPT). Here, L, Q, H, P, and T represent linear, quadratic, hinge, product, and threshold, respectively. The optimal parameter combination was determined based on the corrected Akaike Information Criterion (AICc), and the combination with the lowest delta AICc (delta.AICc = 0) was selected to achieve the best balance between model complexity and goodness of fit. Furthermore, the model predictive performance was evaluated using the mean area under the curve (AUC) (auc.diff.avg). The final optimized settings for MaxEnt were RM = 0.5 and FC = LQHPT.\u003c/p\u003e\n\u003cp\u003eFor the remaining nine models, the optimal parameter settings in the Bigboss options of biomod2 were used,\u0026nbsp;which should give better results than the default set. During model training, 75% of the occurrence data were randomly assigned for model training, while the remaining 25% were used for testing, with ten iterations of random validation. Model accuracy was evaluated using two key metrics: the true skill statistic (TSS) and the area under the receiver operating characteristic curve (ROC). TSS, also known as the Hanssen-Kuiper skill score, is a metric that integrates sensitivity and specificity [46]. The ROC curve plots the true positive rate against the false positive rate across different classification thresholds, and reflects the model\u0026rsquo;s classification ability at various threshold levels. The AUC value close to 1 indicates better model performance [47]. We compared the AUC and TSS values of 10 models based on the simulation results of the modern distribution. Finally, the RF model was selected for subsequent simulations of the species suitable habitat distribution.\u003c/p\u003e\n\u003cp\u003eThe final SDMs predicted the probability (p) of species occurrence within the study area, and this allows for the classification of suitable and unsuitable habitats. Suitable habitats were further categorized into three suitability levels\u0026mdash;low suitable (0.2 \u0026le; p \u0026lt; 0.4), moderately suitable (0.4 \u0026le; p \u0026lt; 0.6), and extremely suitable (p \u0026ge; 0.6) [48].\u003c/p\u003e"},{"header":"3 Results","content":"\u003ch2\u003e3.1 Current distribution\u003c/h2\u003e\n\u003cp\u003eWe used the biomod2 package to compare the accuracy of ten species distribution models (SDMs) (Fig. 4) and their predicted species distribution results (Fig. 5). Except for the MaxEnt model, the average TSS values of the remaining nine models exceeded 0.95. Similarly, with the exception of the ANN, CTA, and MaxEnt models, the average AUC values of the other seven models were above 0.99. Although most models demonstrated high predictive accuracy, there were notable differences in the distribution of suitable habitats. We found that the RF model exhibited both high accuracy and a better alignment with the actual species distribution. Moreover, it provided a more precise identification of low suitable and moderately suitable regions, in contrast to other models that tended to produce overly binary probability classifications (Fig. 5).\u003c/p\u003e\n\u003cp\u003eWe simulated the global distribution probability of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e, but the probability of suitable habitat outside East Asia is extremely low. Therefore, only the results for the East Asian region are presented. Under current climate conditions, the total suitable habitat area is estimated at 413.76 \u0026times; 10⁴ km\u0026sup2;, comprising an extremely suitable area of 177.35 \u0026times; 10⁴ km\u0026sup2;, a moderately suitable area of 63.06 \u0026times; 10⁴ km\u0026sup2;, and a low suitable area of 173.35 \u0026times; 10⁴ km\u0026sup2;. The species suitable habitats are primarily distributed across China and Japan. The extremely suitable areas are concentrated south of the Qinling-Huaihe Line in China and in the southern and eastern regions of Japan. The moderately suitable and low suitable areas are also widely distributed in regions north of the Qinling-Huaihe Line, including Henan and Shandong provinces, as well as Yunnan Province and the Himalayan region, Hainan and Taiwan provinces, Vietnam, the Korean Peninsula, and various parts of Japan (Fig. 6).\u003c/p\u003e\n\u003ch2\u003e3.2 Variable importance assessment and response curves\u003c/h2\u003e\n\u003cp\u003eThe importance of environmental variables (Fig. 7) and the response curves of species distribution models (Fig. 8) were analyzed to identify key factors that influence the variation in suitable habitats. Among the 12 environmental variables used in the models, the most significant factors affecting the suitable habitat distribution of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e were precipitation of warmest quarter (bio18), precipitation of wettest month (bio13), mean diurnal range (bio2), isothermality (bio3), and annual mean temperature (bio1). Notably, precipitation-related factors (bio18, bio13) had a greater influence than temperature-related factors (bio2, bio3, bio1).\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003eThe response curves of different environmental variables with respect to habitat suitability for \u003cem\u003eS. cuneata\u003c/em\u003e revealed that eight variables showed clear optimal values for its growth: annual mean temperature (bio1), mean diurnal range (bio2), isothermality (bio3), temperature annual range (bio7), precipitation of wettest month (bio13), precipitation of driest month (bio14), precipitation of coldest quarter (bio18), and elevation. The most favorable conditions for \u003cem\u003eS. cuneata\u003c/em\u003e growth were: annual mean temperature \u0026asymp; 15℃, mean diurnal range \u0026asymp; 7℃, isothermality \u0026asymp; 30, temperature annual range \u0026asymp; 23℃, precipitation of wettest month \u0026asymp; 400 mm, precipitation of driest month \u0026asymp; 50 mm, precipitation of coldest quarter \u0026asymp; 750 mm, and elevation \u0026asymp; 0\u0026ndash;300 m and 1500 m. Additionally, precipitation seasonality (bio15) and slope exhibited similar trends, where higher values favored \u003cem\u003eS. cuneata\u003c/em\u003e growth. In contrast, the suitability of \u003cem\u003eS. cuneata\u003c/em\u003e decreased with increasing precipitation of coldest quarter (bio19). Aspect showed the least consistent effect on \u003cem\u003eS. cuneata\u003c/em\u003e growth and had the lowest importance among the variables. Overall, \u003cem\u003eS. cuneata\u003c/em\u003e thrives in regions with moderate temperatures, high precipitation, and pronounced precipitation seasonality, particularly at mid- and low-altitude.\u003c/p\u003e\n\u003ch2\u003e3.3 Species distribution under past climatic conditions\u003c/h2\u003e\n\u003cp\u003eWe mapped the fossil sites and found that the historical distribution range of \u003cem\u003eSargentodoxa\u003c/em\u003e was significantly broader than its present range. The fossil records have ages from the Middle Eocene to the Pleistocene, with occurrences in 1) the Eocene and Miocene-Pliocene of North America, 2) the Eocene, Oligocene, Miocene, and Pliocene of Europe, and 3) the Miocene and Pleistocene of Asia [34\u0026ndash;42] (Fig. 9).\u003c/p\u003e\n\u003cp\u003eThe simulation of species distribution during the LGM and MH periods suggests a significant contraction in the suitable habitat compared to the present (Fig. 10). The extremely suitable area completely disappeared, while the moderately suitable area was almost entirely restricted to China, particularly to regions south of the Qinling-Huaihe Line. The suitable habitat was more contracted during the LGM period than in the MH period. During the MH period, the total suitable habitat area was 326.70 \u0026times; 10⁴ km\u0026sup2;, representing a 21.04% reduction compared to the present. The moderately suitable area expanded to 104.40 \u0026times; 10⁴ km\u0026sup2;, with an increase of approximately 65.55%, shifting to regions between the Yangtze-Pearl rivers, which are currently classified as extremely suitable for \u003cem\u003eS. cuneata\u003c/em\u003e. In the LGM period, the total suitable habitat area further decreased to 283.22 \u0026times; 10⁴ km\u0026sup2;, approximately 31.55% less than in the present. The moderately suitable area shrank to 56.03 \u0026times; 10⁴ km\u0026sup2;, with an 11.15% decline, and was primarily confined to the border of Chongqing, Hunan, and Hubei provinces (i.e., the Wuling Mountain Range), as well as the Nanling and Wuyi mountain ranges in China.\u003c/p\u003e\n\u003ch2\u003e3.4 Future suitable distribution under climate change\u003c/h2\u003e\n\u003cp\u003eThe RF model was used to predict the potential suitable habitat of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e under the SSP245 and SSP585 scenarios for the periods 2041\u0026ndash;2060 and 2081\u0026ndash;2100 (Fig. 11b\u0026ndash;e). The SSP245 scenario represents a moderate emission pathway, assuming global efforts to mitigate climate change, while the SSP585 scenario represents a high-emission pathway with limited climate mitigation measures. Overall, under different climate scenarios, the total suitable area is expected to increase, while the area of extremely suitable habitat is projected to decrease. Under the SSP245 scenario, the extent of change remains relatively moderate for both 2041\u0026ndash;2060 and 2081\u0026ndash;2100. The total suitable area is predicted to expand to 441.94 \u0026times; 10⁴ km\u0026sup2; by 2041\u0026ndash;2060 and 451.80 \u0026times; 10⁴ km\u0026sup2; by 2081\u0026ndash;2100 (Table 3). However, the extremely suitable area is projected to shrink to 174.90 \u0026times; 10⁴ km\u0026sup2; and 172.10 \u0026times; 10⁴ km\u0026sup2;, respectively. Under the SSP585 scenario, changes are expected to be minor by 2041\u0026ndash;2060 but become more pronounced by 2081\u0026ndash;2100. The total suitable area will increase to 447.22 \u0026times; 10⁴ km\u0026sup2; by 2041\u0026ndash;2060, while the extremely suitable area will decline to 173.69 \u0026times; 10⁴ km\u0026sup2;. By 2081\u0026ndash;2100, the total suitable area is predicted to expand to 525.11 \u0026times; 10⁴ km\u0026sup2;, marking a 26.91% increase compared to the present. However, the extremely suitable area is expected to decline by 11.44%, shrinking to 157.06 \u0026times; 10⁴ km\u0026sup2;.\u003c/p\u003e\n\u003cp\u003eBy calculating the difference between the future and current distributions of suitable habitat, we identified regions where habitat suitability is expected to increase or decrease under different climate scenarios (Fig. 11f\u0026ndash;i). From the suitable distribution and changes in suitability, all scenarios suggest that suitability in the southern regions will decline, while northern areas will generally experience an increase. The suitability of coastal regions in southern China, most parts of Yunnan, and portions of the Himalayan region is expected to decrease, particularly under the SSP585 scenario by 2081\u0026ndash;2100, where the decline is more pronounced. Conversely, regions north of the Qinling-Huaihe Line, including the North China Plain, Northeast China Plain, Hengduan Mountains, eastern Xizang Province (Fig. 2b), as well as central and northern Japan, are expected to experience an increase in habitat suitability. Notably, under the SSP585 scenario by 2081\u0026ndash;2100, the suitability of the Northeast China Plain will increase significantly.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3.\u003c/strong\u003e The different-level suitable area for the present, past, and future (10\u003csup\u003e4\u003c/sup\u003e km\u003csup\u003e2\u003c/sup\u003e), and the areal differences were calculated compared with the present suitable area (%).\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePeriod\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eExtremely suitable\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModerate suitable\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLow suitable\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 19px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eArea\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChange\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eArea\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChange\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eArea\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChange\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eArea\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChange\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eModern\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e177.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e63.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e173.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e413.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eLGM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e-100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e56.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e-11.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e227.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e31.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e283.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e-31.55\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eMH\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e-100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e104.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e65.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e222.31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e28.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e326.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e-21.04\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eSSP245 2041\u0026ndash;2060\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e174.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e-1.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e81.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e29.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e185.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e6.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e441.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e6.81\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eSSP245 2081\u0026ndash;2100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e172.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e-2.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e87.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e38.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e192.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e11.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e451.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e9.19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eSSP585 2041\u0026ndash;2060\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e173.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e-2.07\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e85.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e34.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e188.44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e8.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e447.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e8.09\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003eSSP585 2081\u0026ndash;2100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e157.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11px;\"\u003e\n \u003cp\u003e-11.44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9px;\"\u003e\n \u003cp\u003e96.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e53.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e271.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e56.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e525.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10px;\"\u003e\n \u003cp\u003e26.91\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eUsing the centroid calculation function in ArcGIS, we analyzed the shifts in the centroid of the extremely suitable area under different SSP scenarios (Fig. 12). Currently, the centroid of the extremely suitable area is located in central-southern Hubei Province (112.94\u0026deg; E, 30.55\u0026deg; N). Under the SSP245 scenario, the centroid is projected to first shift southeastward to southeastern Hubei Province (114.58\u0026deg; E, 30.16\u0026deg; N), with a displacement of 163.22 km, and then move northwestward to 113.85\u0026deg; E, 30.29\u0026deg; N, covering a distance of 71.61 km. These shifts remain within Hubei Province. Under the SSP585 scenario, the centroid is expected to first move eastward to eastern Hubei Province (114.47\u0026deg; E, 30.59\u0026deg; N), with a shift of 146.55 km, and then shift further southeast beyond Hubei Province into northwestern Anhui Province, reaching coordinates of 116.31\u0026deg; E, 29.94\u0026deg; N, after an additional displacement of 190.92 km.\u003c/p\u003e"},{"header":"4 Discussion","content":"\u003ch2\u003e4.1 Model performance evaluation\u003c/h2\u003e\n\u003cp\u003eThis study compared the predictive performance of 10 species distribution models (SDMs) implemented in the biomod2 platform to simulate the current distribution of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e. While all models exhibited high accuracy, their predicted geographical distributions showed noticeable differences. The observed differences in predicted species distribution probabilities may be due to the complex relationships between variables. Furthermore, each model uses different methods to define the probability of species occurrence in relation to environmental variables. Based solely on AUC and TSS values, the MARS and GLM models slightly outperformed the RF model. However, in terms of geographical distribution, RF provided results that more closely matched the actual distribution pattern. The superiority of the RF model may stem from its ability to handle complex, high-dimensional data, particularly its capacity to effectively model nonlinear responses and mitigate the risk of overfitting [49].\u003c/p\u003e\n\u003cp\u003eThe performance of different models varies across different species [48]. Some studies have found that RF achieves the highest accuracy [44, 50], whereas others report that MaxEnt outperforms other models [6, 51]. This highlights the inherent uncertainty across different modeling approaches and underscores the importance of selecting the most appropriate model based on the specific research objectives. Despite the strong performance of RF herein, its potential limitations should be acknowledged. Substantial biases may often arise in predictions generated by a single model under future climate scenarios. Ensemble modeling, which integrates predictions from multiple models, has been proposed as a strategy to enhance overall prediction robustness [52, 53]. However, ensemble models do not always yield the best results, as an optimized single model can sometimes outperform an ensemble approach [15]. Although RF performed well here, uncertainties in future predictions remain a concern, particularly under the complex dynamics of climate change. Therefore, future research could explore the use of ensemble modeling approaches while simultaneously optimizing the parameters of both single and ensemble models based on specific research objectives and data characteristics to achieve more robust and reliable predictions.\u003c/p\u003e\n\u003ch2\u003e4.2 Potential dispersal routes of\u003cem\u003e\u0026nbsp;Sargentodoxa\u003c/em\u003e\u003c/h2\u003e\n\u003cp\u003eBased on available fossil records,\u003cem\u003e\u0026nbsp;Sargentodoxa\u0026nbsp;\u003c/em\u003ehas been distributed exclusively in the Northern Hemisphere, encompassing three major regions (Europe, North America, and Asia). The earliest fossils of \u003cem\u003eSargentodoxa\u003c/em\u003e were discovered along the western coast of North America from the middle Eocene, suggesting that the genus may have originated in North America (Fig. 9, Table 1) [34]. North America is also likely the diversification center for \u003cem\u003eSargentodoxa\u003c/em\u003e. Subsequently, \u003cem\u003eSargentodoxa\u003c/em\u003e spread to Europe and Asia while continuing to disperse across North America.\u003c/p\u003e\n\u003cp\u003eIt is reasonable to hypothesize that \u003cem\u003eSargentodoxa\u003c/em\u003e migrated from North America to Europe via the North Atlantic Land Bridge (NALB). There are two possible routes by which \u003cem\u003eSargentodoxa\u003c/em\u003e may have dispersed to Asia. The first route is a direct expansion from North America to Asia via the Bering Land Bridge (BLB). The second involves an initial spread to Europe, followed by an eastward migration across the Eurasian continent, eventually reaching Japan and southwestern China. However, the route across Eurasia remains highly uncertain due to the lack of fossil evidence. This migration could have occurred through the northern boreotropical migration via Siberia, the southern route through the Kohistan-Ladakh Island Arc (KLIA) [54], or a pathway between these two possibilities. This suggests that the dispersal of \u003cem\u003eSargentodoxa\u003c/em\u003e was most likely facilitated by long-distance dispersal (LDD). Both the NALB and the BLB were critical migration routes for numerous tropical, subtropical, and temperate species across the Northern Hemisphere during the Cenozoic\u0026nbsp;[55, 56]. Extensive studies based on fossil records have suggested that species dispersed from North America to Europe via the NALB through Iceland\u0026nbsp;[57–59] and reached Asia from North America through the BLB [60]. In the future, the discovery of more fossils of \u003cem\u003eSargentodoxa\u003c/em\u003e is expected to clarify its dispersal pathways and provide a more comprehensive understanding of its biogeographical history.\u003c/p\u003e\n\u003ch2\u003e4.3 Identification of glacial refugia\u003c/h2\u003e\n\u003cp\u003eThe cyclical shifts between glacial and interglacial periods since the Quaternary have profoundly influenced modern species distributions and genetic differentiation [61]. Here, species distribution models were used to simulate the suitable habitat of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e during the LGM and MH. The results indicate a significant southward contraction of the species suitable range compared to the present, with suitability levels decreasing. Notably, the extremely suitable area nearly disappeared entirely. During the LGM (21–18 thousand years ago, ka), global temperatures were approximately 7.0 ± 1.0°C lower than pre-industrial levels [62]. The climate in East Asia was particularly harsh, driving tropical and subtropical species southward. As \u003cem\u003eS. cuneata\u003c/em\u003e is an indicator species of warm and humid environments, its growth was severely constrained by low temperatures, leading to a more restricted distribution and reduced suitability. Consequently, compared to the MH period, the suitable habitat during the LGM experienced further contraction, with the moderately suitable area also shrinking.\u003c/p\u003e\n\u003cp\u003eIntegrating fossil records from the Cenozoic era, we infer that the glacial refugia of \u003cem\u003eS. cuneata\u003c/em\u003e were primarily located in the Nanling Mountains, the Wuyi Mountains, and the Wuling Mountains in China. These refugia overlap with those identified for East Asian relict species [19] and coincide with known refugial areas of the \"living fossil\" \u003cem\u003eGinkgo biloba\u003c/em\u003e in Southwest, East, and South China [5, 63]. Among these,the Nanling Mountains (23°37′–27°14′ N) are the largest mountain range in southern China, serving as a natural biogeographical boundary for the subtropical zone. This region is recognized as a biodiversity hotspot, providing refugium for numerous relict species due to its unique topography and ecosystem [64]. The Wuyi Mountains, situated in southeastern China along the border of Jiangxi and Fujian provinces, are characterized by highly complex topography, diverse habitats, and favorable climatic conditions. Due to these factors, the Wuyi Mountains have been widely recognized as a natural gene bank for biological species [65]. This makes their role as a refugium for \u003cem\u003eS. cuneata\u003c/em\u003e unsurprisingly. The Wuling Mountains, a northeast-southwest trending mountain range in central China (27.28°–30.05° N, 107.02°–111.33° E), exhibit complex and diverse vegetation and serve as a biodiversity hotspot for plant species in central China [66, 67]. This region has also been identified as a refugium for many ancient plant species in China [68]. The orientation of mountain ranges is closely related to the direction of species dispersal [69]. The mountain ranges that served as glacial refugia for \u003cem\u003eS. cuneata\u003c/em\u003e generally follow north-south, east-west, and northeast-southwest directions. These mountains facilitated its post-glacial dispersal across southern China and even East Asia.\u003c/p\u003e\n\u003cp\u003eFossil evidence further reveals that \u003cem\u003eSargentodoxa\u003c/em\u003e had a much wider distribution in the Paleogene (~40 Ma) and Neogene (~15 Ma), with a global presence. However, since the Quaternary (~2.58 Ma), its distribution has become restricted to China (Fig. 9, Table 1). This pattern suggests that \u003cem\u003eSargentodoxa\u003c/em\u003e experienced severe geographical contraction during the climatic fluctuations of the Quaternary.\u0026nbsp;The existence of multiple, spatially dispersed glacial refugia likely enabled the species to persist in small populations during harsh glacial-interglacial cycles. These refugia facilitated postglacial range expansions across the Chinese subtropics, gradually shaping its current restricted distribution [27]. Given their critical role in sustaining biodiversity and ecosystem stability, these refugial regions warrant prioritized conservation efforts.\u003c/p\u003e\n\u003ch2\u003e4.4 Impact of climate change on its distribution\u003c/h2\u003e\n\u003cp\u003eThe analysis of species response curves to environmental factors sheds light on the understanding of the appropriate distribution of species under future climate change and the intrinsic reasons for the changes. Analysis of the species response curves to environmental variables revealed that precipitation plays a slightly more critical role than temperature. As a climbing plant, \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e has high water requirements, and sufficient precipitation directly influences its growth and reproductive success. Numerous studies have identified water availability as a key limiting factor for liana growth (e.g., Jiang et al., 2011). Additionally, the response curve of precipitation seasonality (bio15) exhibits a strong linear relationship, indicating that \u003cem\u003eS. cuneata\u003c/em\u003e thrives in environments with pronounced seasonal precipitation variation. This suggests that the species has effectively adapted to the distinct seasonal rainfall patterns associated with the East Asian monsoon climate. The ample precipitation in these regions provides the moist environment necessary for its growth. Furthermore, researchers observed a \"midday depression\" phenomenon in the photosynthesis of \u003cem\u003eS. cuneata\u003c/em\u003e leaves during noon hours of summer, indirectly supporting our finding that the species suitability declines when temperatures exceed a certain threshold [71]. Interestingly, the suitable altitude range for \u003cem\u003eS. cuneata\u003c/em\u003e may exhibit two distinct intervals: a lower altitude range (0–300 m) and a mid-altitude range (around 1500 m). Among these, the lower altitude range appears to be more suitable than the mid-altitude range, although this observation lacks sufficient literature support. Additionally, within the lower altitude range of approximately 0–500 m, habitat suitability decreases with increasing altitude. A possible explanation is that \u003cem\u003eS. cuneata\u003c/em\u003e is more easily discovered at lower altitudes, leading to a higher number of occurrence records compared to those at mid-altitudes, which introduces bias. Future studies should incorporate a more comprehensive set of occurrence points and consider the physiological characteristics of \u003cem\u003eS. cuneata\u003c/em\u003e to further refine the analysis of its suitable altitude range. Overall, \u003cem\u003eS. cuneata\u003c/em\u003e is best adapted to mid-and low-altitude regions with moderate temperatures, abundant precipitation, and distinct precipitation seasonality.\u003c/p\u003e\n\u003cp\u003eThe response curves of environmental factors indirectly confirms that mountainous regions have been, and will continue to be, suitable habitats for \u003cem\u003eS. cuneata\u003c/em\u003e. Mountains provide diverse topographies that offer a wide range of ecological niches, ensuring species survival across various environmental conditions. Moreover, mountain uplift influences atmospheric circulation and provides abundant orographic precipitation. The diversity of soil types and the abundance of water resources in mountainous regions further enhance the availability of essential survival resources for species. In China, the east-west- and northeast-southwest-oriented mountain ranges also block cold air from moving southward. For instance, the climate south of the Qinling-Huaihe Line is warmer and more humid. These unique geomorphological and ecological characteristics make mountains important refugia for numerous species, which play a critical role in maintaining global biodiversity during glacial and interglacial cycles [72, 73]. For example, the Hengduan Mountains in China have preserved a wide range of endemic and relict plant species by providing stable habitats [74]. This highlights the vital role that mountainous refugia play in biodiversity conservation.\u003c/p\u003e\n\u003cp\u003eAccording to IPCC projections, by 2100, global surface temperatures are expected to rise by 2.7 °C under the moderate-emission SSP245 scenario and by 4.4 °C under the high-emission SSP585 scenario [75].\u0026nbsp;Our predictions of future habitat suitability indicate that under the SSP245 scenario, the species suitable habitat will not differ significantly from its current distribution, with a general northward shift. However, under the SSP585 scenario, the suitable range will undergo a more pronounced northward expansion. Climate warming is identified as the primary driver of \u003cem\u003eS. cuneata\u003c/em\u003e northward expansion and southern range contraction.\u003c/p\u003e"},{"header":"5 Conclusion","content":"\u003cp\u003eIn this study, we compared nine machine learning models and the MaxEnt model to assess their predictive performance, ultimately selecting the RF model for simulating the past, present, and future suitable habitats of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e under climate change scenarios. The results indicate that \u003cem\u003eS. cuneata\u003c/em\u003e thrives in mid- and low-altitude regions with moderate temperatures, abundant precipitation, and distinct precipitation seasonality, with its extremely suitable habitats primarily located south of the Qinling-Huaihe Line in China. Precipitation-related factors had a greater influence on distribution than temperature-related factors. During the LGM and MH periods, the species suitable range contracted significantly, with reductions of 130.54 km² and 87.06 km², respectively. The contraction was most pronounced during the LGM, when its suitable habitat was restricted to the Nanling Mountains, Wuyi Mountains, and Wuling Mountains—potential glacial refugia for the species. By 2100, model simulations indicate that under two different scenarios (SSP245 and SSP585), the total area of suitable habitat is expected to expand, with a noticeable trend of northward expansion and a corresponding contraction in the south. While the centroid of highly suitable habitats is expected to shift slightly eastward. As a representative relict plant species in China, \u003cem\u003eS. cuneata\u003c/em\u003e shares similar responses to climate change and glacial refugia with other South Asian relict species. This study will place an important basis for future conservation strategies and climate adaptation measures for relict plants, and thus contribute to biodiversity conservation and sustainable ecosystem management. Future research combining more comprehensive fossil data, genetic analysis, and ecological modeling can further elucidate the evolutionary history, dispersal, and geographic distribution of \u003cem\u003eS. cuneata\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor contributions\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eX.L.:\u003c/strong\u003e Conceptualization, Methodology, Investigation, Data curation, Validation, Visualization, Writing - Original Draft, Writing - Review \u0026amp; Editing. \u003cstrong\u003eH.H.:\u003c/strong\u003e Conceptualization, Methodology, Investigation, Writing - Review \u0026amp; Editing, Supervision, Funding acquisition. \u003cstrong\u003eX.M.:\u003c/strong\u003e Validation, Visualization, Writing - Review \u0026amp; Editing. \u003cstrong\u003eM.L.:\u003c/strong\u003e Validation, Visualization, Writing - Review \u0026amp; Editing.\u003cstrong\u003e\u0026nbsp;Z.Q.:\u003c/strong\u003e Visualization, Writing - Review \u0026amp; Editing.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis work was supported by the Starting Grant for Introduced Talents of Sun Yat-sen University, the Fundamental Research Funds for the Central Universities, Sun Yat-sen University (No. 24qnpy021), and the General Project of Basic and Applied Basic Research of Guangzhou Bureau of Science and Technology (No. 2025A04J4384).\u003c/p\u003e\n\u003ch2\u003eData availability\u003c/h2\u003e\n\u003cp\u003eData is provided within the manuscript or supplementary information files.\u003c/p\u003e\n\u003ch2\u003eClinical trial number\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eEthics approval and consent to participate\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eConsent for publication\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eAcknowledgments\u003c/h2\u003e\n\u003cp\u003eWe thank the National Herbarium Resource Center for providing modern distribution data of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e. We also acknowledge Dr. Renbin Zhu for providing photos of \u003cem\u003eS. cuneata\u003c/em\u003e.\u003c/p\u003e\n\u003ch2\u003eCompeting Interest\u003c/h2\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eFu J, Wen L. Impacts of Quaternary glaciation, geological history and geography on animal species history in continental East Asia: A phylogeographic review. Mol Ecol. 2023;32:4497\u0026ndash;514.\u003c/li\u003e\n\u003cli\u003eGuo W, Yang Y, Zhang X, Chen J, Wu S, Yang J, et al. Genomic divergence between two sister \u003cem\u003eMedicago\u003c/em\u003e species triggered by the quaternary climatic oscillations on the Qinghai\u0026ndash;Tibet plateau and northern China. Mol Ecol. 2023;32:3118\u0026ndash;32.\u003c/li\u003e\n\u003cli\u003eYin Q-Y, Fan Q, Li P, Truong D, Zhao W-Y, Zhou R-C, et al. Neogene and Quaternary climate changes shaped the lineage differentiation and demographic history of \u003cem\u003eFokienia hodginsii\u003c/em\u003e (Cupressaceae s.l.), a Tertiary relict in East Asia. J Syst Evol. 2021;59:1081\u0026ndash;99.\u003c/li\u003e\n\u003cli\u003eWiens JJ, Zelinka J. How many species will earth lose to climate change? Glob Chang Biol. 2024;30:e17125.\u003c/li\u003e\n\u003cli\u003eWang L, Liu J, Liu J, Wei H, Fang Y, Wang D, et al. Revealing the long-term trend of the global-scale \u003cem\u003eGinkgo biloba \u003c/em\u003edistribution and the impact of future climate change based on the ensemble modeling. Biodivers Conserv. 2023;32:2077\u0026ndash;100.\u003c/li\u003e\n\u003cli\u003eKang Y, Lin F, Yin J, Han Y, Zhu M, Guo Y, et al. Projected distribution patterns of \u003cem\u003eAlpinia officinarum\u003c/em\u003e in China under future climate scenarios: Insights from optimized Maxent and Biomod2 models. Front Plant Sci. 2025;16.\u003c/li\u003e\n\u003cli\u003eQi X, Chen C, Comes HP, Sakaguchi S, Liu Y, Tanaka N, et al. Molecular data and ecological niche modelling reveal a highly dynamic evolutionary history of the East Asian Tertiary relict \u003cem\u003eCercidiphyllum\u003c/em\u003e (Cercidiphyllaceae). New Phytol. 2012;196:617\u0026ndash;30.\u003c/li\u003e\n\u003cli\u003eTang CQ, Dong Y-F, Herrando-Moraira S, Matsui T, Ohashi H, He L-Y, et al. Potential effects of climate change on geographic distribution of the Tertiary relict tree species \u003cem\u003eDavidia involucrata\u003c/em\u003e in China. Sci Rep. 2017;7:43822.\u003c/li\u003e\n\u003cli\u003eZhao G, Cui X, Sun J, Li T, Wang Q, Ye X, et al. Analysis of the distribution pattern of Chinese Ziziphus jujuba under climate change based on optimized biomod2 and MaxEnt models. Ecological Indicators. 2021;132:108256.\u003c/li\u003e\n\u003cli\u003eElith J, Leathwick JR. Species distribution models: Ecological explanation and prediction across space and time. Annu Rev Ecol Evol Syst. 2009;40 Volume 40, 2009:677\u0026ndash;97.\u003c/li\u003e\n\u003cli\u003ePhillips SJ, Anderson RP, Schapire RE. Maximum entropy modeling of species geographic distributions. Ecol Modell. 2006;190:231\u0026ndash;59.\u003c/li\u003e\n\u003cli\u003eBreiman L. Random forests. Mach Learn. 2001;45:5\u0026ndash;32.\u003c/li\u003e\n\u003cli\u003eNelder JA, Wedderburn RWM. Generalized linear models. J R Stat Soc Ser A. 1972;135:370\u0026ndash;84.\u003c/li\u003e\n\u003cli\u003eHastie T, Tibshirani R. Generalized additive models: Some applications. J Am Stat Assoc. 1987;82:371\u0026ndash;86.\u003c/li\u003e\n\u003cli\u003eHao T, Elith J, Lahoz-Monfort JJ, Guillera-Arroita G. Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models. Ecography. 2020;43:549\u0026ndash;58.\u003c/li\u003e\n\u003cli\u003eLi X, Wang Y. Applying various algorithms for species distribution modelling. Integr Zool. 2013;8:124\u0026ndash;35.\u003c/li\u003e\n\u003cli\u003eM\u0026eacute;dail F, Diadema K. Glacial refugia influence plant diversity patterns in the Mediterranean Basin. J Biogeogr. 2009;36:1333\u0026ndash;45.\u003c/li\u003e\n\u003cli\u003eHampe A, Rodr\u0026iacute;guez‐S\u0026aacute;nchez F, Dobrowski S, Hu FS, Gavin DG. Climate refugia: From the Last Glacial Maximum to the twenty‐first century. New Phytol. 2013;197:16\u0026ndash;8.\u003c/li\u003e\n\u003cli\u003eTang CQ, Matsui T, Ohashi H, Dong Y-F, Momohara A, Herrando-Moraira S, et al. Identifying long-term stable refugia for relict plant species in East Asia. Nat Commun. 2018;9:4488.\u003c/li\u003e\n\u003cli\u003ePetit RJ, Aguinagalde I, de Beaulieu J-L, Bittkau C, Brewer S, Cheddadi R, et al. Glacial refugia: Hotspots but not melting pots of genetic diversity. Science. 2003;300:1563\u0026ndash;5.\u003c/li\u003e\n\u003cli\u003eSingh PB, Mainali K, Jiang Z, Thapa A, Subedi N, Awan MN, et al. Projected distribution and climate refugia of endangered Kashmir musk deer \u003cem\u003eMoschus cupreus\u003c/em\u003e in greater Himalaya, South Asia. Sci Rep. 2020;10:1511.\u003c/li\u003e\n\u003cli\u003eChen DZ, Tatemi S. Lardizabalaceae. Flora of China. 2001;6:440\u0026ndash;54.\u003c/li\u003e\n\u003cli\u003eZhao X, Ke H, Yu H. Studies on pharmacological effect and clinical application of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e. Guiding Journal of Traditional Chinese Medicine and Pharmacy. 2014;20:41\u0026ndash;3.\u003c/li\u003e\n\u003cli\u003eZhang W, Sun C, Zhou S, Zhao W, Wang L, Sheng L, et al. Recent advances in chemistry and bioactivity of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e. J Ethnopharmacol. 2021;270:113840.\u003c/li\u003e\n\u003cli\u003eWang Y, Zhang B, Liu S, Xu E, Wang Z. The traditional herb \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e alleviates DSS-induced colitis by attenuating epithelial barrier damage via blocking necroptotic signaling. J Ethnopharmacol. 2024;319:117373.\u003c/li\u003e\n\u003cli\u003eXu F, Yu P, Wu H, Liu M, Liu H, Zeng Q, et al. Aqueous extract of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e alleviates ulcerative colitis and its associated liver injuries in mice through the modulation of intestinal flora and related metabolites. Front Microbiol. 2024;15.\u003c/li\u003e\n\u003cli\u003eTian S, Lei S-Q, Hu W, Deng L-L, Li B, Meng Q-L, et al. Repeated range expansions and inter-/postglacial recolonization routes of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e (oliv.) rehd. et wils. (Lardizabalaceae) in subtropical China revealed by chloroplast phylogeography. Mol Phylogenet Evol. 2015;85:238\u0026ndash;46.\u003c/li\u003e\n\u003cli\u003eManchester SR, Chen Z-D, Lu A-M, Uemura K. Eastern Asian endemic seed plant genera and their paleogeographic history throughout the Northern Hemisphere. Journal of Systematics and Evolution. 2009;47:1\u0026ndash;42.\u003c/li\u003e\n\u003cli\u003eZhou Z, Arata M. Fossil history of some endemic seed plants of east asiaand its phytogeographical significance. Acta Bot Yunnan. 2005;27:449\u0026ndash;70.\u003c/li\u003e\n\u003cli\u003eGBIF.org. GBIF occurrence downloaded on 23rd december 2024. Available at https://doi.org/1015468/dl.zsphbx. 2024.\u003c/li\u003e\n\u003cli\u003eZizka A, Silvestro D, Andermann T, Azevedo J, Duarte Ritter C, Edler D, et al. CoordinateCleaner: Standardized cleaning of occurrence records from biological collection databases. Methods Ecol Evol. 2019;10:744\u0026ndash;51.\u003c/li\u003e\n\u003cli\u003eWarren DL, Matzke NJ, Cardillo M, Baumgartner JB, Beaumont LJ, Turelli M, et al. ENMTools 1.0: An R package for comparative ecological biogeography. Ecography. 2021;44:504\u0026ndash;11.\u003c/li\u003e\n\u003cli\u003eThuiller W, Georges D, Engler R, Breiner F, Georges MD, Thuiller CW. Package \u0026lsquo;biomod2.\u0026rsquo; Species distribution modeling within an ensemble forecasting framework. 2016;10:1600-0587.2008.\u003c/li\u003e\n\u003cli\u003eManchester SR. Biogeographical relationships of north American Tertiary floras. Ann Mo Bot Gard. 1999;86:472\u0026ndash;522.\u003c/li\u003e\n\u003cli\u003eSong Z. Late Cenozoic palyno-flora from Zhaotong, Yunnan. Proceedings of the Nanjing Institute of Geology and Palaeontology, Chinese Academy of Sciences. 1988;0:1\u0026ndash;108.\u003c/li\u003e\n\u003cli\u003eMai HD. Die mittelmioz\u0026auml;nen und obermioz\u0026auml;nen floren aus der meuroer und raunoer folge in der lausitz. Teil II: Dicotyledonen. Palaeontographica Abteilung B. 2001;257:35\u0026ndash;174.\u003c/li\u003e\n\u003cli\u003eTiffney BH. Fruits and seeds of the Tertiary Brandon Lignite. Vii. \u003cem\u003eSargentodoxa \u003c/em\u003e(Sargentodoxaceae). Am J Bot. 1993;80:517\u0026ndash;23.\u003c/li\u003e\n\u003cli\u003eMcNair D, Stults D, Axsmith B, Alford M, Starnes J. Preliminary investigation of a diverse megafossil floral assemblage from the middle Miocene of southern Mississippi, USA. Palaeontol Electronica. 2019;22.2.40A:1\u0026ndash;30.\u003c/li\u003e\n\u003cli\u003eMomohara A. Change of paleovegetation caused by topographic change in and around a sedimentary basin of the Upper Miocene Tokiguchi Porcelain Clay Formation, central Japan. Geoscience Report of the Shimane University. 2001;20:49.\u003c/li\u003e\n\u003cli\u003eMead JI, Schubert BW, Wallace SC, Swift SL. Helodermatid lizard from the Mio-Pliocene oak-hickory forest of Tennessee, eastern USA, and a review of monstersaurian osteoderms. Acta Palaeontol Polonica. 2012;57:111\u0026ndash;21.\u003c/li\u003e\n\u003cli\u003eGeissert F, Gregor HJ, Mai DH. Die \u0026ldquo;saugbaggerflora\u0026rdquo;: Eine frucht- und samenflora aus dem grenzbereich mioz\u0026auml;n-plioz\u0026auml;n von sessenheim im elsass (frankreich). Forschungen aus den Naturwissenschaften; 1990.\u003c/li\u003e\n\u003cli\u003eMartinetto E. Studies on some exotic elements of the Pliocene floras of Italy. Palaeontographica Abteilung B. 2001;:149\u0026ndash;66.\u003c/li\u003e\n\u003cli\u003eZhang M-Z, Xu Z, Han Y, Guo W. Evaluation of CMIP6 models toward dynamical downscaling over 14 CORDEX domains. Clim Dyn. 2024;62:4475\u0026ndash;89.\u003c/li\u003e\n\u003cli\u003eZhao Z, Xiao N, Shen M, Li J. Comparison between optimized MaxEnt and random forest modeling in predicting potential distribution: A case study with\u003cem\u003e Quasipaa boulengeri\u003c/em\u003e in China. Sci Total Environ. 2022;842:156867.\u003c/li\u003e\n\u003cli\u003eKass JM, Muscarella R, Galante PJ, Bohl CL, Pinilla‐Buitrago GE, Boria RA, et al. ENMeval 2.0: Redesigned for customizable and reproducible modeling of species\u0026rsquo; niches and distributions. Methods Ecol Evol. 2021;12:1602\u0026ndash;8.\u003c/li\u003e\n\u003cli\u003eAllouche O, Tsoar A, Kadmon R. Assessing the accuracy of species distribution models: Prevalence, kappa and the true skill statistic (TSS). J Appl Ecol. 2006;43:1223\u0026ndash;32.\u003c/li\u003e\n\u003cli\u003eShabani F, Kumar L, Ahmadi M. Assessing accuracy methods of species distribution models: AUC, specificity, sensitivity and the true skill statistic. Global Journal of Human-Social Science: B Geography, Geo-Sciences, Environmental Science \u0026amp; Disaster Management. 2018;18.\u003c/li\u003e\n\u003cli\u003eWang P, Luo W, Zhang Q, Han S, Jin Z, Liu J, et al. Assessing the impact of climate change on three populus species in China: Distribution patterns and implications. Glob Ecol Conserv. 2024;50:e02853.\u003c/li\u003e\n\u003cli\u003eValavi R, Elith J, Lahoz‐Monfort JJ, Guillera‐Arroita G. Modelling species presence‐only data with random forests. Ecography. 2021;44:1731\u0026ndash;42.\u003c/li\u003e\n\u003cli\u003eLi C, Luo G, Yue C, Zhang L, Duan Y, Liu Y, et al. Distribution patterns and potential suitable habitat prediction of \u003cem\u003eCeracris kiangsu\u003c/em\u003e (Orthoptera: Arcypteridae) under climate change - a case study of China and Southeast Asia. Sci Rep. 2024;14:20580.\u003c/li\u003e\n\u003cli\u003eCao G, Yuan X, Shu Q, Gao Y, Wu T, Xiao C, et al. Prediction of the potentially suitable areas of \u003cem\u003eEucommia ulmoides\u003c/em\u003e Oliver in China under climate change based on optimized Biomod2 and MaxEnt models. Front Plant Sci. 2024;15.\u003c/li\u003e\n\u003cli\u003eWang D, Shi C, Alamgir K, Kwon S, Pan L, Zhu Y, et al. Global assessment of the distribution and conservation status of a key medicinal plant (\u003cem\u003eArtemisia annua\u003c/em\u003e L.): The roles of climate and anthropogenic activities. Sci Total Environ. 2022;821:153378.\u003c/li\u003e\n\u003cli\u003eWu Y, Shen J, Deane DC, Yu H, Yu F, Wang X, et al. Future extreme climate events threaten alpine and subalpine woody plants in China. Earth\u0026rsquo;s Future. 2025;13:e2024EF005147.\u003c/li\u003e\n\u003cli\u003eGao Y, Song A, Cai W-J, Spicer RA, Zhang R, Liu J, et al. Tibetan Plateau palm fossils prove the Kohistan-Iadakh Island Arc is a floristic steppingstone between Gondwana and Laurasia. Rev Palaeobot Palynol. 2025;334:105255.\u003c/li\u003e\n\u003cli\u003eHopkins DM. Cenozoic history of the Bering Land Bridge. Science. 1959;129:1519\u0026ndash;28.\u003c/li\u003e\n\u003cli\u003eTiffney BH. The Eocene North Atlantic Land Bridge: Its importance in Tertiary and modern phytogeography of the Northern Hemisphere. J Arnold Arbor. 1985;66:243\u0026ndash;73.\u003c/li\u003e\n\u003cli\u003eDenk T, Gr\u0026iacute;msson F, Zetter R, S\u0026iacute;monarson LA. The biogeographic history of Iceland \u0026ndash; the North Atlantic Land Bridge revisited. In: Denk T, Grimsson F, Zetter R, S\u0026iacute;monarson LA, editors. Late Cainozoic Floras of Iceland: 15 Million Years of Vegetation and Climate History in the Northern North Atlantic. Dordrecht: Springer Netherlands; 2011. p. 647\u0026ndash;68.\u003c/li\u003e\n\u003cli\u003eJia L-B, Manchester SR, Su T, Xing Y-W, Chen W-Y, Huang Y-J, et al. First occurrence of \u003cem\u003eCedrelospermum \u003c/em\u003e(Ulmaceae) in Asia and its biogeographic implications. J Plant Res. 2015;128:747\u0026ndash;61.\u003c/li\u003e\n\u003cli\u003eJiang Y, Gao M, Meng Y, Wen J, Ge X-J, Nie Z-L. The importance of the North Atlantic land bridges and eastern Asia in the post-Boreotropical biogeography of the Northern Hemisphere as revealed from the poison ivy genus (\u003cem\u003eToxicodendron\u003c/em\u003e, Anacardiaceae). Mol Phylogenet Evol. 2019;139:106561.\u003c/li\u003e\n\u003cli\u003eWen J, Nie Z, Ickert‐Bond SM. Intercontinental disjunctions between eastern Asia and western North America in vascular plants highlight the biogeographic importance of the Bering land bridge from late Cretaceous to Neogene. J Syst Evol. 2016;54:469\u0026ndash;90.\u003c/li\u003e\n\u003cli\u003eHewitt G. The genetic legacy of the Quaternary ice ages. Nature. 2000;405:907\u0026ndash;13.\u003c/li\u003e\n\u003cli\u003eOsman MB, Tierney JE, Zhu J, Tardif R, Hakim GJ, King J, et al. Globally resolved surface temperatures since the Last Glacial Maximum. Nature. 2021;599:239\u0026ndash;44.\u003c/li\u003e\n\u003cli\u003eZhao Y-P, Fan G, Yin P-P, Sun S, Li N, Hong X, et al. Resequencing 545 ginkgo genomes across the world reveals the evolutionary history of the living fossil. Nat Commun. 2019;10:4201.\u003c/li\u003e\n\u003cli\u003eTian S, Kou Y, Zhang Z, Yuan L, Li D, L\u0026oacute;pez-Pujol J, et al. Phylogeography of \u003cem\u003eEomecon chionantha\u003c/em\u003e in subtropical China: The dual roles of the Nanling Mountains as a glacial refugium and a dispersal corridor. BMC Evol Biol. 2018;18:20.\u003c/li\u003e\n\u003cli\u003eChen L, Cao X, Pan T, Lei P, Zeng L, Li M, et al. Conservation status and prioritization of rare and endangered plants in Jiangxi Wuyishan National Nature Reserve. Journal of Nanjing Forestry University (Natural Sciences Edition). 2024;48:39.\u003c/li\u003e\n\u003cli\u003eSun Z, Yang L, Kong H, Kang M, Wang J. Phylogeographical patterns match the floristic subdivisions: The diversification history of a widespread herb in subtropical China. Ann Bot. 2024;134:1263\u0026ndash;76.\u003c/li\u003e\n\u003cli\u003eWang C, Zhou T, Qin Y, Zhou G, Fei Y, Xu Y, et al. Wuling mountains function as a corridor for woody plant species exchange between northern and southern central China. Front Ecol Evol. 2022;10.\u003c/li\u003e\n\u003cli\u003eQi C, Yu X, Zheng Z, Yin G. Notes on endemic seed plants in the floristic region of central China. Journal of Central-South Forestry College. 1998;18:1\u0026ndash;4.\u003c/li\u003e\n\u003cli\u003eXiao Y, Li X-J, Jiang X-L, Li C, Li X-P, Li W-P, et al. Spatial genetic patterns and distribution dynamics of \u003cem\u003eBegonia grandis \u003c/em\u003e(Begoniaceae), a widespread herbaceous species in China. Front Plant Sci. 2023;14.\u003c/li\u003e\n\u003cli\u003eJiang H, Zhou G-Y, Huang Y-H, Liu S-Z, Tang X-L. Photosynthetic characteristics of canopy-dwelling vines in lower subtropical evergreen broad-leaved forest and response to environmental factors: Photosynthetic characteristics of canopy-dwelling vines in lower subtropical evergreen broad-leaved forest and response to environmental factors. Chin J Plant Ecol. 2011;35:567\u0026ndash;76.\u003c/li\u003e\n\u003cli\u003eJin Z, Ke S, Zhong Z. Studies on characteristics of leaves growth and photosynthesis physioecology of \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e. Bulletin of Botanical Research. 2002;22:184\u0026ndash;9.\u003c/li\u003e\n\u003cli\u003eAntonelli A, Kissling WD, Flantua SGA, Berm\u0026uacute;dez MA, Mulch A, Muellner-Riehl AN, et al. Geological and climatic influences on mountain biodiversity. Nature Geosci. 2018;11:718\u0026ndash;25.\u003c/li\u003e\n\u003cli\u003eHoorn C, Mosbrugger V, Mulch A, Antonelli A. Biodiversity from mountain building. Nature Geosci. 2013;6:154\u0026ndash;154.\u003c/li\u003e\n\u003cli\u003eSun H, Zhang J, Deng T, Boufford DE. Origins and evolution of plant diversity in the Hengduan Mountains, China. Plant Divers. 2017;39:161\u0026ndash;6.\u003c/li\u003e\n\u003cli\u003eIntergovernmental Panel on Climate Change (IPCC). Climate change 2021 \u0026ndash; the physical science basis: Working group I contribution to the sixth assessment report of the intergovernmental panel on climate change. 1st edition. Cambridge University Press; 2021.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Lardizabalaceae, Habitat suitability, Glacial refugia, Random Forest model, Machine learning, Biogeography, Climate change","lastPublishedDoi":"10.21203/rs.3.rs-6516979/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6516979/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGlobal climate change is a critical factor influencing biodiversity and ecosystem stability by altering the suitable habitats of many species. \u003cem\u003eSargentodoxa cuneata\u003c/em\u003e, is an endemic and relict plant species in China. Identifying its suitable habitats across different periods and glacial refugia helps explain how \u003cem\u003eS. cuneata\u003c/em\u003e survived Quaternary climate fluctuations, which is crucial for informing its future conservation. However, long-term tracking of its distribution and systematic description of biogeographical evolution remains scarce. Here, we compare ten species distribution models to assess their predictive performance. Ultimately, we apply a random forest model to simulate the suitable habitats of \u003cem\u003eS. cuneata\u003c/em\u003e under past, present, and future climate scenarios and integrate fossil records to analyze its biogeographical history. We find that \u003cem\u003eS. cuneata\u003c/em\u003e is currently distributed primarily south of the Qinling-Huaihe Line in China, particularly in mid- and low-altitude mountainous regions with abundant precipitation and moderate temperatures. During the Last Glacial Maximum (LGM, about 22,000 years ago) and Mid-Holocene (MH, about 6,000 years ago), its suitable habitat contracted significantly, with extremely suitable areas nearly disappearing due to colder climate. Glacial refugia are identified in three mountain ranges within Central and South China. Model simulations under two different climate scenarios suggest that while the total suitable habitat of \u003cem\u003eS. cuneata\u003c/em\u003e may expand, extremely suitable areas could decline, with a northward expansion and southern contraction. This study will provide insights into the long-term impact of climate change on relict plant species and contribute to a better understanding of the evolutionary history of East Asian flora.\u003c/p\u003e","manuscriptTitle":"The past, present and future distribution of Sargentodoxa Rehder \u0026amp; E.H.Wilson: Perspectives from fossil record and species distribution models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-30 11:08:51","doi":"10.21203/rs.3.rs-6516979/v1","editorialEvents":[{"type":"communityComments","content":1}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"2119a47e-6fac-4bfa-80d1-fc1fa4e8e238","owner":[],"postedDate":"April 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-05-07T04:53:25+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-30 11:08:51","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6516979","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6516979","identity":"rs-6516979","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}