Since the first proposal of the MaxEnt model in 2004, researchers have clearly emphasized that high-quality species occurrence data were needed for high-quality statistical analysis (Phillips et al., 2004). Studies have shown that MaxEnt performs well for species with few occurrence records and is more suitable for predicting the distribution of rare species (Jha et al., 2022). Therefore, MaxEnt was selected in this study to model and predict the future distribution of E. brevicornum. Species distribution data were primarily obtained from the Global Biodiversity Information Facility (GBIF.org (8 July 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.2ktupd), Chinese Virtual Herbarium (http://www.cvh.ac.cn), and keyword literature searches. The longitude and latitude distribution data of this species were determined by using Google Maps (http://ditu.google.cn/). In order to ensure data quality and reduce the impact of data duplication and redundancy on the results, duplicate points within each grid cell (10 km × 10 km) were removed (Zhang et al., 2018b), resulting in a total of 92 distribution points.

The model first selected 19 bioclimatic variables from the WorldClim global climate database (https://worldclim.org/) ( Supplementary Table S1 ). To eliminate multicollinearity among these variables, MaxEnt v3.4.1 was initially used to calculate the contribution rates of the 19 environmental variables ( Supplementary Table S2 ). Then, Pearson correlation analysis was conducted using SPSS 25.0 to exclude environmental factors with an absolute correlation coefficient greater than or equal to 0.8 (Feng et al., 2019; Jayasinghe and Kumar, 2019), resulting in the selection of six bioclimatic variables ( Supplementary Table S3 ). Since various soil components and microorganisms can influence plant growth and distribution, incorporating soil and topographic factors into the species distribution model was necessary (Jiawen et al., 2022; Shen et al., 2024). Therefore, topographic variables, including elevation (alt), slope, and aspect, were obtained from the Resource and Environmental Science and Data Center of the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (https://www.resdc.cn/). Soil variables, including soil pH, topsoil sand content (T-sand), topsoil organic carbon content (T-C), reference soil depth (ref), and soil texture classification (t-usda-te), were sourced from the World Soil Database (https://gaez.fao.org/pages/hwsd). Additionally, the highest monthly mean UV-B radiation (UV-B3) was retrieved from the gIUV database (http://www.ufz.de/gluv/), and the human footprint index (hf) was acquired from the Center for International Earth Science Information Network (CIESIN) (http://www.ciesin.org/). To comprehensively investigate the effects of soil and topographic factors on E. brevicornu, these variables, along with the six previously selected bioclimatic factors, were all incorporated into MaxEnt v3.4.1 for final modeling ( Table 1 ). All environmental factor data were from the 2023 release.

The model was developed based on the Shared Socioeconomic Pathways (SSP) of the Coupled Model Intercomparison Project Phase 6 (CMIP6), proposed by the Intergovernmental Panel on Climate Change (IPCC) (Eyring et al., 2016). Following previous studies, three commonly used SSP scenarios—SSP1-2.6, SSP3-7.0, and SSP5-8.5—were selected for modeling from a total of 23 scenarios (Xu et al., 2024).

During the modeling process, 25% of the data were randomly selected for testing, while the remaining 75% were used as the training set. This procedure was repeated ten times to establish the predictive model. Subsequently, the “kuenm” package in R was used to optimize the regularization multiplier (RM) and feature combination (FC) parameters (Zeng et al., 2016; Kass et al., 2021). The specific procedure was as follows: First, five feature combinations (FCs) and their corresponding regularization multipliers (RMs) were set, generating five candidate models. Then, the “kuenm” package was used to evaluate these models based on the selected parameter combinations. The optimal model was chosen from the candidate models based on two criteria: an omission rate (OR) of <5% and a Delta AICc (Akaike Information Criterion, AIC) value of <2. Finally, among the optimal models, the one with the smallest Delta AICc value was selected as the optimized model, determining its regularization multiplier and feature combination parameters.

The prediction performance indicators of the MaxEnt model include the area under the ROC curve (AUC), the Kappa coefficient (KAPPA), and the True Skill Statistic (TSS). The model’s predictive accuracy is directly proportional to these evaluation metrics. When the value falls between 0.7 and 0.85, it indicates that the model performs well; a value between 0.85 and 1 suggests that the model has excellent performance (Yang et al., 2022; Zhang et al., 2024a).

The distribution data generated by MaxEnt within China was extracted using ArcGIS 10.8 software, and the climate suitability of E. brevicornu was studied in depth based on the existence probability generated by MaxEnt. The suitability was classified into four levels based on the existence probability: highly suitable (0.6-1), suitable (0.3-0.6), less suitable (0.1-0.3), and unsuitable (0-0.1) (Zhang et al., 2018b).

In the ten trials, the AUC value of the E. brevicornum distribution model was 0.949 ( Figure 1 ), the TSS value was 0.860, and the KAPPA value was 0.786 ( Table 2 ). Although the TSS and KAPPA values are lower than the AUC, they still indicate that the model provides good prediction accuracy for the distribution of E. brevicornum. After screening, 6 out of 19 bioclimatic variables were included in the modeling, with a cumulative contribution rate of 74.7%, namely: minimum temperature of the coldest month (bio6) (29.1%), annual precipitation (bio12) (17.2%), precipitation seasonality (bio15) (14.3%), temperature seasonality (bio4) (9.4%), isothermality (bio3) (3.2%), and mean temperature of the wettest quarter (bio8) (1.5%). Because the model needed to consider factors with relatively high contribution rates (Wei et al., 2023), bio3 and bio8, with low contribution rates, were not considered key environmental factors in this work. In the final model, based on the training gains of different climatic factors analyzed using the jackknife method, the top 4 climatic factors in terms of training gain scores were human footprint (hf), bio6, bio12, and Mean UV-B of Highest Month (UV-B3) ( Figure 2 ). Considering both the contribution rate ranking and the training gain ranking, the above-mentioned 6 environmental factors (bio6, bio4, bio12, bio15, hf, UV-B3) were analyzed as key environmental factors affecting the distribution of E. brevicornu.

The MaxEnt model was used to construct the most suitable habitat prediction map for E. brevicornu ( Figure 3 ), The model’s predictions closely matched the current distribution of E. brevicornum. Under the current climatic conditions, the high suitability score zone for E. brevicornu was mainly concentrated in the area of 25.13°-39.50°N, 102.46°-118.13°E, primarily located in the Loess Plateau. In addition, certain suitable areas were also identified in the Sichuan Basin, Yunnan-Guizhou Plateau, and some parts of the southeastern coast.

According to the predicted area of E. brevicornu under the current climate conditions ( Supplementary Table S4 ; Table 3 ), Gansu, Shanxi, Shaanxi, and Henan provinces had a relatively high proportion of high suitability areas, with their respective area proportions of 4.12%, 7.06%, 11.75%, and 4.84% ( Table 3 ). The total areas of high, medium, and low suitable areas in these provinces was 6 × 10³ km², 19 × 10³ km², and 25 × 10³ km², respectively, with a total area of 44 × 10³ km² ( Table 3 ).

The suitable range of E. brevicornu in the 2050s and 2090s, based on the predictions of the three climate change scenarios SSP1-2.6, SSP2-4.5, and SSP5-8.5, can be seen in Figure 4 , with all ranges increasing. It is expected that in the 2050s, the area of high suitability for the species would mainly be concentrated in the Loess Plateau and northern Sichuan, with some suitable areas also present in the Yungui Plateau, southern hilly areas, and the North China Plain. Under the SSP1-2.6 scenario, the area of high suitability increased the most, with significant increases in areas such as Guizhou, Hunan, Hubei, Jiangxi, and Shandong. Under the SSP2-4.5 scenario, the area of high suitability in regions such as the Loess Plateau increased compared to SSP1-2.6, but decreased in other areas. The increase in the area of high suitability under the SSP5-8.5 scenario was less pronounced, with unclear differences compared to the current distribution. From the 2050s to the 2090s, the area of high suitability decreased under the SSP1-2.6 and SSP2-4.5 scenarios, approaching the area under the SSP5-8.5 scenario, but still increased compared to the current high suitability area.

According to the predictions, by the 2050s and 2090s, the suitable range for E. brevicornu was expected to be wider than present ( Table 4 ). In the 2050s, the predicted increase in the suitable range for E. brevicornu under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios was 396.94%, 416.33%, and 17.49%, respectively. Among these, the SSP2-4.5 scenario showed the largest increase in the high suitability range for E. brevicornu, followed by the SSP1-2.6 scenario. In the 2090s, the predicted increase in the suitable range for E. brevicornu under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios was 21.06%, 10.61%, and 14.87%, respectively. Compared to the 2050s, the area of high suitability will decrease in the 2090s.

According to the response curve of E. brevicornu’s probability of occurrence in relation to environmental variables in the MaxEnt model ( Figure 5 ), the suitable range (optimal value) of environmental variables for E. brevicornu distribution was determined ( Table 5 ). The seasonal variation coefficient of air temperature ranged from 735.52 to 1013.95°C (mean: 783.64°C), the lowest temperature in the coldest month ranged from -12.88 to -1.67°C (mean: -5.54°C), annual precipitation ranged from 503.96 to 948.71 mm (mean: 582.11 mm), the seasonal variation in precipitation ranged from 64.22 to 90.59 mm (mean: 72.71 mm), human disturbance ranged from 22.31% to 50.00% (mean: 35.44%), and the maximum monthly average UV radiation ranged from 4496.05 to 5111.58 kJ/m² (mean: 4761.76 kJ/m²). Within the suitable range, the probability of occurrence increased as all environmental variable values approached the optimal value, but began to decrease after surpassing the optimum. It was worth noting that the probability of occurrence of E. brevicornu showed a secondary increase trend after exceeding 5658.76 kJ/m² for the maximum monthly average UV radiation, although the probability was not very high.

Under the current climatic conditions, the centroid of the high-suitability area for E. brevicornu was located in Foping County, Xi’an City, Shaanxi Province ( Figure 6 ). In three future climate scenarios, the centroid of the high-suitability area shifted to varying degrees, but the overall direction of the shift was southeast, with the shifted centroids all remaining in Shaanxi Province ( Figure 6 ). Under the SSP1-2.6 scenario, by the 2050s, the centroid moved to Yunxi County, Shiyan City (32.95°N, 110.20°E), with a distance of 124.20 km²; by the 2090s, the centroid moved to Zhen’an County, Shangluo City (33.43°N, 109.49°E), with a distance of 112.91 km². Under the SSP2-4.5 scenario, by the 2050s, the centroid moved to Xunyang County, Ankang City (33.08°N, 109.50°E), with a distance of 31.84 km²; by the 2090s, the centroid moved to Zhen’an County, Shangluo City (33.30°N, 109.27°E), with a distance of 96.83 km². Under the SSP5-8.5 scenario, by the 2050s, the centroid moved to Yunxi County, Shiyan City (33.06°N, 109.84°E), with a distance of 54.22 km²; by the 2090s, the centroid moved to Zhen’an County, Shangluo City, Hengyang County (33.28°N, 109.34°E), with a distance of 104.14 km². By the 2050s, the centroids were located in Shiyan City and Ankang City, respectively; by the 2090s, they were both located in Shangluo City. Protected areas for E. brevicornu could be established with these cities as the centers.

Epimedium brevicornu is widely used as a medicinal plant in both traditional medicine and modern drug preparation. A total of 29 bioclimatic variables were selected as the primary environmental factors, and the MaxEnt model was employed to simulate the potential suitable distribution of E. brevicornu in China. According to the results of the MaxEnt model, E. brevicornu is currently distributed primarily within the range of 25.13°–39.50°N, 102.46°–118.13°E in China, as well as in the region at 47.29°N, 130.25°E. The most suitable growth areas predominantly include the southern parts of Gansu, Shaanxi, Shanxi, western Henan, northern Sichuan, and northern Hunan provinces in China, which aligns with previous research findings (Ma et al., 2011). This indicates that the Loess Plateau region is more suitable for the survival of E. brevicornum compared to other areas. In the future, climate change will have a significant impact on the distribution of E. brevicornum. Under the SSP1-2.6 and SSP2-4.5 scenarios, the highly suitable growth areas in the 2050s will expand significantly, primarily in the Loess Plateau, further confirming the species’ preference for distribution in this region. However, by the 2090s, the extent of the highly suitable areas will decrease compared to the 2050s, with the area approaching that under the SSP5-8.5 scenario. In the future, highly suitable areas in all scenarios will increase compared to the current situation.

The habitat of a species is often closely related to local bioclimatic, topographical, and soil factors (He et al., 2021). In this study, key environmental variables affecting the occurrence probability of E. brevicornu were analyzed, and the corresponding response curves were generated. The results demonstrated that the probability of occurrence of E. brevicornu varied with changes in key environmental factors. Temperature emerged as the most significant variable affecting its distribution, particularly the seasonal variation coefficient of temperature and the lowest temperature in the coldest month. Precipitation was a secondary factor, with the most significant impacts observed for annual precipitation and seasonal variation in precipitation. In addition, human footprint and UV-B radiation were also found to influence the distribution of E. brevicornu.

Temperature plays a dominant role in plant growth and development (Gao et al., 2022) and influences plant geographical distribution (Ding and Yang, 2022). Studies have shown that differences in species’ global distribution are largely determined by their cold tolerance levels (Araújo et al., 2013). In this study, E. brevicornum was found to be most suitable at a minimum temperature of -5.38°C, where the probability of occurrence was highest. Low temperatures can affect seed dormancy and germination, thus influencing the plant’s growth cycle (Inouye, 2020). E. brevicornum seeds exhibit dormancy, and the optimal temperature range for dormancy is 2-6°C, which is consistent with the modeling results (Li et al., 2024a). This suggests that the temperature limitation for E. brevicornum is primarily during the seed dormancy period. Furthermore, some studies have shown that the growth of E. brevicornum increases significantly with a 2°C temperature rise compared to a 5°C rise (Xuemei et al., 2016). This indicates that E. brevicornum is better suited to moderate temperature changes and struggles to adapt to more extreme temperature shifts. The temperature rise is higher under the SSP5-8.5 scenario, while it is smaller under the SSP1-2.6 and SSP2-4.6 scenarios. As a result, the highly suitable areas in future low-emission scenarios will increase more significantly, while the changes in high-emission scenarios will be smaller. Precipitation changes can affect species distribution and have been shown to maintain relationships between different plant species (Ding and Yang, 2022). When precipitation increases excessively, the productivity of perennial herbaceous plants decreases (Gherardi and Sala, 2015), likely because excess moisture causes root rot and promotes the growth of pests and diseases (Duan et al., 2022). The most suitable conditions for E. brevicornum growth occur when the annual precipitation is 582.11 mm and the seasonal precipitation variation is 72.36 mm, which is similar to the precipitation levels in the Loess Plateau (Wu et al., 2024). Research has found that natural drought in the soil after rainfall during the harvest period can enhance the quality of related plants (Zhang et al., 2024b), which could explain why E. brevicornum is widely distributed in the Loess Plateau. Human footprint can alter terrestrial ecosystems, and the extent and magnitude of these changes vary over time and space (Watson and Venter, 2019), with environmental changes caused by human activity being complex and diverse (Kennedy et al., 2019). Although the inclusion of human footprint in SDM models is relatively recent, human activity remains one of the most important factors influencing plant distribution (Amat et al., 2013; Gallardo et al., 2015) and has different effects on various plants (Bai et al., 2024), promoting the distribution of some high-altitude plants (Amat et al., 2013). In this study, it was found that as the human footprint increased, the probability of E. brevicornum distribution also increased. Although this probability decreased after reaching 35.44%, it remained relatively high, suggesting that human activity may promote the distribution of E. brevicornum. UV-B radiation can cause changes in plant metabolism, physiology, and development (Liu et al., 2024). UV-B stress reduces the leaf area of E. brevicornum but increases the accumulation of flavonoid compounds (Li et al., 2024b). The model indicated that when the highest monthly UV-B radiation ranged between 4587.88 and 4769.53 kJ/m², it had a positive impact. However, as UV-B radiation exceeded 4769.53 kJ/m² and approached 5658.76 kJ/m², the probability of occurrence decreased with increasing UV-B radiation. When the radiation exceeded 5658.76 kJ/m², the occurrence probability showed a secondary increase, although the peak probability was lower. This may be because plants adapt to UV-B stress to protect themselves (Shi and Liu, 2021). This has been confirmed in other high-altitude plant species (Shi et al., 2022), but further research is needed for E. brevicornum, which is widely distributed in the Loess Plateau.

The wild populations of E. brevicornum are under threat, making conservation efforts essential. The primary task in conserving E. brevicornum should be protecting its native ecological environment. Based on modeling results, conservation efforts should focus on the main distribution area of E. brevicornum, the Loess Plateau, taking into full consideration the key environmental factors that influence its growth. Secondly, expanding artificial cultivation is important. At present, the cultivation of medicinal plants by replicating the living conditions of wild plants has already been implemented (He et al., 2021). Studies have shown that the centers of high suitability areas for E. brevicornum are in Shangluo, Shiyan, and Ankang cities, as well as their surrounding regions. Therefore, efforts should be focused on conducting surveys, collecting resources, and protecting these areas, which should also be promoted as important bases for planting and development. Finally, human excavation activities should be restricted to protect E. brevicornum. Frequent human activities can limit the ability of protected areas to maintain ecosystem integrity (Xia et al., 2021). As a medicinal plant, E. brevicornum is subject to exploitation of its wild resources, which threatens its wild populations. Remote sensing can monitor human activities and facilitates the establishment and assessment of protected areas (Xia et al., 2024). Therefore, in the future wild conservation areas of E. brevicornum, remote sensing and other geospatial technologies could be integrated in practical applications to monitor and restrict human activities, which would be more beneficial for the conservation of wild E. brevicornum resources.

There were still some limitations in the model prediction process. First, the role of the model was limited to prediction, and the extent of future environmental changes may differ from the three scenarios used in the model (Meinshausen et al., 2022). Moreover, rising temperatures are expected to cause extreme climatic events in the Loess Plateau region (Ta et al., 2022). Since not all environmental variables could be taken into account (Xu et al., 2022), prediction errors are inevitable, and the future distribution of E. brevicornu may also differ from the predicted patterns. In addition, interspecific interactions, such as competition, predation, and mutualisms, were not fully considered (Van Ee et al., 2022). These factors were not included in the modeling and were not thoroughly investigated in this study, but they may potentially affect the distribution of E. brevicornum. Therefore, in practical applications, the model should be used only as a reference. After identifying suitable habitats, effective conservation of the species requires integrating more local environmental factors.