By consulting online databases such as the Chinese Virtual Herbarium (http://www.cvh.ac.cn/) and the NSII-China National Specimen Resource Platform (http://www.nsii.org.cn/), a total of 303 records were collected nationwide and their distribution information was also obtained ( Supplementary Table S1 ). Then, duplicate coordinate data and samples with unclear geographic distribution location information were removed. According to the reported method (Wu et al., 2024), the coordinate information of sample points with clear geographical locations was determined using baidu coordinate picker (https://api.map.baidu.com/lbsapi/getpoint/index.html), and 89 samples of P. kadsura were finally obtained ( Supplementary Table S2 ). When utilizing this tool, if the provided location information was not accurate beyond the district or county level, automatic identification of coordinates became unattainable. In such cases, manual positioning was required. When manually locating an area, it was crucial to confine within the boundaries of the county. Otherwise, there might be significant deviations. Additionally, to reduce model overfitting caused by sampling bias, neighborhood analysis in ArcGIS 10.4.1 was used to set a buffer zone with a radius of 10 km, and one distribution point was randomly retained within a range of 20 km, eventually resulting in 65 valid distribution points ( Supplementary Table S3 ; Figure 1 ). The species name, longitude, and latitude of these points were applied for subsequent analysis (Zhang et al., 2019).

Nineteen climate variables were obtained from the World Climate Database (http://www.worldclim.org) using current (1970-2000) climate data as the baseline, selecting the Last Glacial Maximum (LGM) and the Mid-Holocene (MH) for past climate data, as well as different scenarios for future climate (2041-2060, 2081-2100). Future climate data were determined based on the Shared Socioeconomic Pathways (SSPs) models released by the Sixth Coupled Model Intercomparison Project (CMIP6), with SSP126 (low emission scenario) and SSP585 (high emission scenario) reflecting the most optimistic and pessimistic greenhouse gas emission scenarios for the future, respectively (Riahi et al., 2017). Meanwhile, eleven soil variables and three topographic variables were obtained from the Food and Agriculture Organization of the United Nations World Soil Database (http://www.fao.org/soils-portal/data-hub/en/) and the WorldClim website (https://www.worldclim.org/) (Fan et al., 2022; Ouyang et al., 2022). A total of 33 environmental variables were applied to evaluate the impact on distribution of P. kadsura, and the most dominant environmental variables were found out after eliminating strongly correlation variables that are relatively minor ( Table 1 ).

In order to reduce the high correlation and multicollinearity among environmental variables that cause model overfitting and ensure the accuracy of the prediction, this study used SPSS 26.0 software to perform Spearman correlation analysis on the above environmental variables ( Supplementary Figure S1 ). When the correlation coefficient of two environmental variables was greater than |0.8|, the variables with small contribution rates were eliminated, thus minimizing the bias fitting of the MaxEnt mode (Yang et al., 2022). Ultimately, 18 environmental variables were retained to construct the prediction model for P. kadsura, including 8 climatic variables (Bio17, Bio12, Bio7, Bio2, Bio3, Bio18, Bio1, Bio15), 7 soil variables (t_ph_h2o, s_clay, s_oc, t_sand, s_caco3, s_ph_h2o, t_oc), and 3 topographic variables (aspect, elev, slope).

The distribution data of P. kadsura and effective environmental variables were imported into the MaxEnt software (V3.4.3) to predict its potential suitable habitat distribution. The following modeling parameters were used: sampling method was bootstrap, output format was logistic, and 75% of the distribution points were randomly selected as the training set, with the remaining 25% of the distribution points as the test set. For each training partition, after 106 iterations and 10 times model repetition, the average value of the calculations was taken as the final result of the model prediction (Huang et al., 2023). This study selected the area under the Receiver Operating Characteristic (ROC) curve (AUC) to evaluate the accuracy of the model prediction (Xu et al., 2020). At the same time, the Jackknife method was used to analyze the impact of each environmental variable on the distribution of P. kadsura and to plot the response curves of key environmental variables. In addition, in species distribution modeling, the Maximum Test Sensitivity Plus Specificity Logistic Threshold (MTSPS) was used as the dividing line between suitable and unsuitable areas, which is considered simple and effective (Aidoo et al., 2022; Huang R. et al., 2022). With reference to the methods of Yang et al. (2023) and Wu et al. (2024), we used MTSPS to classify their potential suitable habitats into the following four levels the MTSPS was used to divide its potential suitable habitat into the following four levels: unsuitable habitat (0~MTSPS), low suitability habitat (MTSPS~0.3), medium suitability habitat (0.3~0.5), and high suitability habitat (0.5~1), and the area of different suitability habitats was calculated (Yan et al., 2020).

By simulating and predicting the distribution area of the P. kadsura through MaxEnt software, the average ROC curve of 10 calculation results was finally obtained after 10 loops. The AUC value ranged from 0 to 1, and the closer it approached 1, the more accurate the prediction result of the model was. When AUC was more than 0.9, the model prediction was excellent (Ouyang et al., 2022). As shown in Figure 2 , the average training value of the ROC curve in this study was 0.969, indicating that the construction of this model had a very high accuracy and could be used to study the potential suitable habitat of the P. kadsura.

To characterize the effects of various environmental variables on the construction results of the prediction model, we used the MaxEnt model to analyze the contribution rates and permutation importance of 18 environmental variables separately. The percent contribution represented the percentage of the impact of climatic factor on the model after all variables are considered, while permutation importance indicated the degree of impact on the model after the factor has been replaced (Hou J. et al., 2023). As shown in Table 2 , Bio17 (precipitation of driest quarter) had the highest contribution rate at 52.0%, followed by Bio12 (annual precipitation) at 21.9%. The contribution rates of Bio7 (temperature annual range), slope, Bio2 (mean diurnal range), Bio3 (isothermality), aspect, and elev (elevation) were 5.8%, 4.5%, 2.8%, 2.8%, 2.6%, and 2.1% respectively. The contribution rates of the remaining environmental variables were all below 2.0%. Among them, Bio2, Bio7, and elev had relatively high confidence importance values of 38.8%, 17.1%, and 10.7% respectively, indicating a strong dependence of the model on these three variables (Wang E. et al., 2024).

To characterize the importance of various environmental variables on the distribution of P. kadsura, we used the jackknife test to examine the impact of dominant environmental factors on the suitable distribution area of P. kadsura in China ( Figure 3 ). The results indicated that Bio12, Bio17, and Bio2 had the greatest impact on the distribution of P. kadsura, suggesting that these three environmental variables contain more effective information compared to others (Deng et al., 2024). On the whole, the dominant environmental variables influencing the distribution of P. kadsura were precipitation of Bio17, Bio12, Bio2, and Bio7. Therefore, it could be inferred that temperature and precipitation are key factors affecting the distribution of P. kadsura.

Then, based on the response curves of the key environmental variables derived above, the relationship between the distribution probability of P. kadsura and the environmental variables can be determined. When the distribution probability of P. kadsura was greater than 0.5, the corresponding environmental variable values were favorable for the growth of P. kadsura. The response curves ( Figures 4A–D ) showed that the value ranges (and optimal values) of the key environmental variables that limited the distribution of P. kadsura were: Bio17 100.68-274.48 mm (153.24 mm), Bio12 1194.10-3898.20 mm (2190.12 mm), Bio7 7.82-28.00°C (12.66°C), Bio2 3.65-8.06°C (6.42°C). The distribution probability raised with the increase in the values of key environmental variables before the optimal values, and decreased with the increase in the environmental factor values after the optimal values.

According to the “Flora Reipulicae Popularis Sinicae”, P. kadsura was distributed along the coastal areas of China, especially in the provinces of Fujian and Zhejiang, growing in low-altitude forests, climbing on trees or rocks (Flora of China Editorial Committee of Chinese Academy of Sciences, 1982). The distribution records in the NSII-China National Specimen Resource Platform showed that P. kadsura is mainly distributed in Guangxi (129 distribution points), Guangdong (94 distribution points), Fujian (77 distribution points), Zhejiang (69 distribution points), Guizhou (42 distribution points), Taiwan (35 distribution points), Yunnan (22 distribution points), Sichuan (16 distribution points), Jiangxi (15 distribution points), and Hainan (12 distribution points), with sporadic distribution in other provinces. As shown in Figure 5 , the white areas represented unsuitable zones of P. kadsura, the green areas represented low suitability zones, the yellow areas represented medium suitability zones, and red represents high suitability zones. The main distribution range of P. kadsura was between 105° E - 121° E and 18° N - 30° N, including medium and high suitability zones, with a total suitable area of 51.74 × 10⁴ km², accounting for 5.39% of China’s land surface area, while the high suitability zone was accounting for only 22.32% of the total suitable area ( Table 3 ). Currently, the distribution of the total suitable area for P. kadsura was relatively concentrated, mainly located in the coastal areas of East and South China, with rare distribution in inland regions, and none as it moved further north, which was highly consistent with the natural distribution area recorded in the “Flora Reipulicae Popularis Sinicae”. Among them, the high suitability zones were mainly distributed in Taiwan, Guangxi, Guangdong, Zhejiang, and Fujian provinces. The medium suitability zones were distributed around the high suitability zones, mainly covering Jiangxi, Hainan, southern Anhui, southern Hunan, southeastern Guizhou, and southeastern Yunnan. The low suitability zone area was 38.49 × 10⁴ km², accounting for 4.01% of China’s land surface area ( Table 3 ). The unsuitable zones were mostly located in the northern and southwestern regions of China, with large areas in Henan, Hubei, northern Jiangsu, and northern Anhui.

This study selected 6 periods to predict the potential distribution of P. kadsura in China. Based on the prediction results of the MaxEnt model, habitat suitability distribution maps of P. kadsura under two scenarios (SSP126, SSP585) in the 2050s and 2090s were obtained. As shown in Figures 6A, B , from the Last Glacial Maximum (LGM) to the Mid-Holocene (MH), P. kadsura had no suitable habitat, with a total suitable area of 0. From MH to the present, the suitable area increased to the maximum, with the current total suitable area being 51.74 × 10⁴ km². The high suitable area covered 11.55 × 10⁴ km², accounting for 1.20% of China’s land surface area ( Table 3 ). This indicated that the current climate was more suitable for the survival of P. kadsura, while during the LGM and MH periods, P. kadsura could not survive, which might be related to the cold climate during the LGM and the unstable climate during the MH period.

From the present to the future, the distribution range of P. kadsura would be reduced to varying degrees, showing a trend of initial decrease followed by an increase under the SSP126 scenario. However, compared to the current distribution range, there would still be a certain degree of reduction ( Figures 6C–F ). The future suitable habitats of P. kadsura shrinking and expanding were shown in Table 3 and Figure 7 . Under the SSP126 scenario, the total suitable area from 2041 to 2060 was 14.03 × 10⁴ km², a decrease of 72.89% compared to the current climate scenario. The high suitable area decreased by 78.09%, while the low and medium suitable areas decreased by 17.72% and 71.39%, respectively. From 2081 to 2100, the total suitable area was 49.48 × 10⁴ km², a decrease of 4.37% compared to the current climate scenario. The low suitable area increased by 10.32%, while the high and medium suitable areas decreased by 14.69% and 1.40%, respectively. Under the SSP585 scenario, the total suitable area from 2041 to 2060 was 20.77 × 10⁴ km², a decrease of 59.85% compared to the current climate scenario. The low suitable area increased by 13.08%, while the high and medium suitable areas decreased by 74.86% and 55.54%, respectively. From 2081 to 2100, the total suitable area was 5.56 × 10⁴ km², a decrease of 89.26% compared to the current climate scenario. The high, medium, and low suitable areas decreased by 87.95%, 89.63%, and 51.55%, respectively.

Current and future climatic scenarios were modeled for the potential distribution of high-value and low-yield medicinal plant P. kadsura in the coastal areas of China. Precipitation contributed the most to the model scores, up to 73.9%, of which the precipitation of driest quarter (Bio17) accounted for 52%, and the annual precipitation (Bio12) accounted for 21.9%. Following was the temperature range that contributed 8.6% to the prediction scores, of which temperature annual range (Bio7) accounted for 5.8%, and the mean diurnal range (Bio2) accounted for 2.8%. These high-contribution variables were similar to those for Eremochloa ophiuroides (Xu et al., 2024). Specifically, P. kadsura maintained good growth efficiency when Bio17 was between 100.68-274.48 mm, Bio12 was between 1194.10-3898.20 mm, Bio7 was between 7.82-28.00°C, and Bio2 was between 3.65-8.06°C. Plants generally required sufficient water to meet transpiration needs and maintain normal physiological functions. When some researchers cultivated P. kadsuram, the precipitation of the selected locations as 1750.00-1800.00 mm (Jiang, 2017) and 1662.00 mm (Li, 2015), which was in line with the predicted range of 1194.10-3898.20 mm for Bio12. Additionally, the geographic distribution of P. kadsura was mainly in the coastal areas of eastern and southern China, which were characterized by a subtropical monsoon climate with distinct seasons and warm, humid conditions, also aligned well with the prediction results.

Besides precipitation, temperature played a significant role in the formation and distribution of plants. With the increase of temperature, the stomatal opening on the plant surface enlarged, the plant transpiration and respiratory rate increased significantly, and eventually lead to substantial losses of water, thereby inhibiting growth (Zhu et al., 2023). The primary temperature influence factors for P. kadsura were the Bio7 and Bio2, achieving optimal growth efficiency at 7.82-28.00°C and 3.65-8.06°C, respectively. This indicated that P. kadsura was not suitable for areas with large temperature differences and was not found in northern regions with significant temperature variations. Moreover, appropriate diurnal temperature variation can promote plant growth (Sun et al., 2000). In this study, the suitable value of Bio2 for the P. kadsura growth was 3.65-8.06°C, and exceeding this range was detrimental to its growth. Furthermore, numerous studies have also shown that precipitation and temperature were crucial variables affecting species distribution. For instance, Chang et al. (2020) studied the impact of climate change on the potential distribution of Anabasis aphylla in Northwestern China. Li et al. (2022) predicted the potential suitable areas for Glycyrrhiza uralensis, and Jiang et al. (2023) explored the potential suitable areas for Panicum milliaceum under climate change, all concluded that temperature and precipitation were major factors influencing the potential distribution of plants.

Therefore, in the future protection and cultivation of P. kadsura, the influence of temperature and precipitation should be fully considered. The results of this study can provide information for the suitable habitat of P. kadsura, but further practical exploration and summary are needed for subsequent practical applications.

As global climate continues to be warm and intensified, accompanied by frequent extreme events, the suitable habitats distribution of many species will be reduced, and the habitat fragmentation will be serious (Barbarossa et al., 2021). Sudden changes in the living environment will affect the migratory ability of species. If a species has weak migratory ability and its distribution speed is slower than the rate of climate change, it will not be able to adapt to the climate change quickly, making it easy for sensitive and ecologically poorly adaptable species to decline in distribution or become extinct (Zhang et al., 2019). Therefore, understanding the distribution of species’ survival under climate change is of great importance for assessing the impact of climate change on species and formulating conservation measures. Additionally, genetic diversity of species should also be fully considered. Potential suitable habitat simulations during different periods indicate that climate change significantly affects the species. The predicted results of this study show that P. kadsura did not have any distribution during the LGM and MH periods, possibly because the LGM was the most recent extremely cold period, with approximately 24% of the global land covered by ice and frequent extreme cold events (Zhan et al., 2022), which did not meet the survival conditions of P. kadsura. During the MH period, the climate was warmer and more humid than the present, with significant climate fluctuations. P. kadsura is a perennial vine, the instability and abrupt change of climate had a great impact on its growth in the next year, making this period also unsuitable for its survival. From the MH period to the present, P. kadsura transitioned from no distribution to having the largest total suitable habitat area, indicating that the current climate conditions favor the growth of P. kadsura.

Compared to the present, the future high-temperature environment caused by carbon emissions showed an overall shrinking trend in the suitable habitat of P. kadsura. Under the low-emission SSP126 scenario, the future suitable habitat area of P. kadsura fluctuates significantly, with a sharp reduction in 2050S, shrinking by 72.89% compared to the current total suitable habitat area, and generally retreating to the southeastern coastal areas. This indicated that the environmental conditions during this period were not suitable for the growth of P. kadsura. In this scenario, the distribution of P. kadsura in the 2090S period was more optimistic than in the 2050S period but still reduced compared to the present. The total suitable habitat area was only reduced by 4.35% compared to the present, with an insignificant reduction degree, but the high suitability area was reduced by 14.69%. This indicated that the climate conditions in the high suitability area during this period did not provide a better growth environment for P. kadsura compared to the present, thus limiting its growth. Under the high-emission SSP585 scenario, the suitable habitat area of P. kadsura shrunk rapidly in the future, with a reduction of 89.25% by the 2090S period compared to the present. This indicated that under the high-emission scenario, P. kadsura cannot adapt to the changing climate environment, experiencing severe growth limitations, and might face endangerment and extinction in the future. The study by He and Ding (2023) showed that compared to the high-emission scenario, the low-emission scenario had a greater possibility of reducing future climate risks. In the face of climate change, the distribution of suitable habitats for plants will respond to varying degrees. However, in the context of continuous warming, the future distribution of P. kadsura is unfavorable, with a reduction in total suitable habitat area. This is consistent with research findings for Alternanthera philoxeroides (Yan et al., 2020), Dipteronia sinensis (Huang Y. et al., 2022) and Entodon challenger (Cong et al., 2023).

Many researches only use climate variables to predict suitable habitats for species, this study applies climate (Zhang et al., 2021; Zuo et al., 2022), soil and terrain as environmental variables, which can improve the accuracy of the suitable habitats prediction for P. kadsura. However, there are many models developed and ensembled to analyze potential suitable habitats. Kunwar et al. (2023) used ensemble model to predict the distribution of seven medicinal plant species of Nepal. Subedi et al. (2023) predicted the distribution of the endangered Maple Leaf oak (Quercus acerifolia) using an integrated model. MaxEnt model stands out for our study because it requires fewer samples and has the advantages of accurate prediction. The prediction of our model may not exactly match actual developments, because only a single MaxEnt model is applied, and the SSP model is based on assumptions of future conditions, rather than direct observation. In addition, model predictions alone are not sufficient to confirm claims about the evolutionary history or origin of P. kadsura. In the following study, we will try to use more models, increase sample quantity, apply R language method, and analyze actual distribution point samples to improve the accuracy of prediction. And consider incorporating phylogenetic evidence or fossil data to strengthen the inferences about the past distributions and evolutionary history of P. kadsura. However, it is important to note that despite some limitations, this study still provide reference value for the sustainable development and utilization of P. kadsura, as well as add literature support for applying species distribution models to assess the effects of climate change on the future distribution of species.

According to the above analysis, we need to implement protective measures for the resources with high-value and low-yield resources. First of all, for the areas that having distribution records or areas identified as high suitability, it is essential to clarify their specific geographic locations and growth patterns, carry out the continuous monitoring about the surrounding habitat and growth conditions, and strengthen the personalized protection of the environmental conditions. Furthermore, it’s essential to identify suitable locations for cultivation and conservation, and then conduct the necessary transplantation, cultivation, and breeding activities to establish a strong foundation for the responsible development and exploitation of P. kadsura.