According to “Flora of China”, C. multiflorus is widely distributed in China, with its natural range extending from the northern part of Altay Prefecture in Xinjiang Uygur Autonomous Region to the southwestern part of Yunnan Province, and from the western end of Wanda Mountains in the eastern part of Heilongjiang Province to the western end of Bortala Mongolian Autonomous Prefecture, Xinjiang Uygur Autonomous Region. The natural distribution range of C. multiflorus is approximately 80°~130° E and 23°~48° N. To avoid overlooking potential suitable areas for C. multiflorus within China and to provide reference for future seed introduction, the study area was expanded to 72°~136° E and 0°~55° N ( Figure 2 ).

The distribution data of C. multiflorus in this study were obtained from field surveys and databases. We conducted field surveys of natural populations of C. multiflorus in the Ningxia Hui Autonomous Region from March to July 2021, collecting 84 distribution records. Other distribution records were obtained by searching specimen and literature databases, including Global Biodiversity Information Facility (GBIF) ¹ , National Specimen Infrastructure (NSII) ² , Chinese Virtual Herbarium (CVH) ³ , Plant Species Information System ⁴ , China National Knowledge Infrastructure (CNKI) and Web of Science (SCI), as well as native flora, collecting 2,057 distribution records. A total of 2,141 distribution records were obtained in this study. To reduce model errors caused by clustering effects, we imported the C. multiflorus distribution point data into ArcGIS, removed artificially planted distribution points, ambiguous records, duplicate distribution points, and conducted buffer zone analysis (Li et al., 2016). Only one distribution point was retained within each 1 km × 1 km grid, resulting in a final set of 539 valid samples.

The provincial administrative map of China used in this study was sourced from the National Basic Geographic Information System Database of China. Contemporary and future climate data [2050s (2041-2060), 2090s (2081-2100)] were obtained using BCC-CMN2-MR in Worldclim 2.1, including SSP1-RCP2.6 (SSP126), SSP2-RCP4.5 (SSP245), and SSP5-RCP8.5 (SSP585). ‘SSP’ represents Shared Socioeconomic Pathway, with SSP126, SSP245, and SSP585 representing low, middle, and high greenhouse gas emission scenarios respectively, while ‘RCP’ represents Representative Concentration Pathways (Popp et al., 2016). This study included seven climate backgrounds (Current, SSP126-2050s, SSP126-2090s, SSP245-2050s, SSP245-2090s, SSP585-2050s, SSP585-2090s), each containing 19 environmental variables with a spatial resolution of 2.5’. To avoid overfitting of the model, Pearson correlation analysis was used in SPSS 25.0 software to screen environmental variables. Environmental variables with |r|< 0.8 were retained, while those with |r| > 0.8 were selected based on their close relationship with C. multiflorus growth (Tang et al., 2021; Xu D. et al., 2019). Finally, 8 environmental variables were selected, as shown in Table 1 .

When optimizing the MaxEnt model by calling the ENMeval package in R software, adjusting the optimal values of the regularization multiplier (RM) and feature categories (FC) can significantly improve the prediction accuracy (Phillips and Dudík, 2008; Cobos et al., 2019; Kass et al., 2021). In this study, we used the EMNeval 2.0 program package in R v4.3.1 for parameter optimization, setting RM in the range of 0.1 to 6 with an interval of 0.5, resulting in a total of 13 regulation multipliers. We also used 10 FC: H, L, LQ, LQH, LQHP, and LQHPT, where L represents linear, Q represents quadratic, H represents hinge, P represents product, and T represents threshold. The ENMeval program package tested the above 78 parameter combinations, and the model fitting degree and complexity were determined using the delta AICc model based on the Akaike information criterion (Kuiper and Hoijtink, 2011). After determining the RM and FC parameter combinations, the collected C. multiflorus sample distribution points and environmental variables were imported into the MaxEnt software. 75% of the sample data was randomly selected as the training set, and 25% as the test set. The maximum number of iterations was set to 1000, and the performance of Bootstrap was repeated 10 times. Other MaxEnt parameters were set to default values.

The simulation accuracy was evaluated using the Receiver Operating Characteristic curve (ROC curve). The area enclosed by the ROC curve and the horizontal axis is the AUC (Area Under Curve) value. The range of AUC values is 0~1, with higher values indicating better prediction performance. Generally speaking, an AUC value less than 0.7 indicates poor prediction performance, 0.7~0.8 indicates relatively accurate, 0.8~0.9 very accurate, and 0.9~1 extremely accurate (Swets, 1988; Araújo et al., 2005; Mahmoodi et al., 2022).

The simulation result files were imported into ArcGIS software for classification of suitability areas. This study employed Jenks’ natural breaks grading method to divide the habitat of C. multiflorus into four levels: unsuitable area (0~0.1), low suitable area (0.1~0.3), medium suitable area (0.3~0.5), and highly suitable area (0.5~1) (Ren et al., 2020). Furthermore, using the ‘Overlay Analysis’ tool in ArcGIS 10.5, the contemporary suitable habitat of C. multiflorus was served as a reference to compare and define the following three scenarios based on changes in the range of C. multiflorus: (1) Retained suitable area: areas suitable in both contemporary and future periods; (2) Increased suitable area: areas currently unsuitable but suitable in the future; (3) Lost suitable area: areas currently suitable but unsuitable in the future. Area statistics and visual representation were conducted using ArcGIS 10.5 (Khwarahm, 2020).

Importance value is a quantitative index used to characterize the status of species in the community. Great importance value means the species is dominant in the community (Feroz et al., 2008). After field investigation (supplementary 1), the main tree species with important value greater than 1% were selected for niche research. Niche overlap reflects the similarity in resource utilization among plants. A higher niche overlap index suggests greater similarity in resource requirements and increased competition (Yuan et al., 2021). Niche breadth was measured by Levins index (B_L ) and Shannon index (B_S ), niche overlap was measured by Pianka index (O_ik ). The calculation formulas were as follows in Equations 1–4 (Fang et al., 2009; Liu R. H. et al., 2018; Liu et al., 2006):

In the above formulas, P_ij is the ratio of utilization of species i on resource position j to the total utilization of all resource positions, P_ik is in the same way. r represents the total number of resource bits. The value domains of B_L and B_S are [1,r] and [0,1nr], respectively. O_ik is the niche overlap index between species i and k, ranging from 0 to 1.

When FC=LQHPT and RM=2.5, the ΔAICc value reached its minimum (ΔAICc = 0) ( Figure 3 ), indicating that the model with this parameter combination was the optimal one. Under these parameters, the result of MaxEnt prediction showed that the value of AUC was 0.875 ( Figure 4 ).

According to Table 2 , the top five contributing environmental variables were Annual precipitation (Bio12, 32.9%), Mean air temperature of the coldest quarter (Bio11, 25.0%), Min air temperature of the coldest month (Bio6, 13.8%), Precipitation of the coldest quarter (Bio19, 11.4%), and Precipitation of the driest month (Bio14, 6.2%), totaling 89.3%. The permutation importance ranked the top five environmental variables as Annual precipitation (Bio12, 37.6%), Mean air temperature of the coldest quarter (Bio11, 20.0%), Precipitation of the coldest quarter (Bio19, 14.4%), Precipitation of the driest month (Bio14, 11.0%), and Min air temperature of the coldest month (Bio6, 8.1%), totaling 91.1%.

The Jackknife test ( Figure 5 ) showed that when individual environmental variables were sequentially excluded, the model’s regularized training gain, test gain, and AUC values decreased most notably for Annual precipitation (Bio12), Precipitation of the driest month (Bio15), and Mean air temperature of the wettest quarter (Bio8). When using only individual variables, the top three with the highest regularized training gain and AUC value gains included Min air temperature of the coldest month (Bio6), Mean air temperature of the coldest quarter (Bio11), and Mean air temperature of driest quarter (Bio9), while the top three with the highest test gain included Min air temperature of the coldest month (Bio6), Mean air temperature of the coldest quarter (Bio11), and Annual precipitation (Bio12).

The response curves of main environmental variables ( Figure 6 ) indicated that C. multiflorus could thrive in environments with Annual precipitation (Bio12) greater than 150 mm, Min air temperature of the coldest month (Bio6) higher than -42.5°C, and Mean air temperature of the coldest quarter (Bio11) higher than -20°C.

The prediction results of the MaxEnt model ( Figure 7 ) showed that the current potential suitable area of C. multiflorus was much larger than the actual collection distribution range ( Figures 1 , 7 ). The total suitable area of C. multiflorus during this period was 411.20 × 10⁴ km², accounting for 42.83% of the total area of the country ( Table 3 ). The highly suitable areas were mainly distributed in a few areas of Shaanxi, Shanxi, southern Ningxia, southern Gansu, northern Sichuan, Xinjiang Uygur Autonomous Region, and Tibet Autonomous Region, concentrated in arid, semi-arid, and semi-humid areas on both sides of the isohyet of 400 mm ( Figure 7 ), indicating that C. multiflorus had strong adaptability and a wide distribution. The highly, medium and low suitable areas accounted for 24.82%, 36.19% and 38.99% of the total suitable area, respectively.

From the results of predicting suitable areas for C. multiflorus under future climate scenarios ( Figure 8 ; Table 3 ), it can be seen that in the 2050s, the total suitable area of C. multiflorus under SSP126, SSP245, and SSP585 slightly increased. It is worth noting that the highly and medium suitable area of C. multiflorus under the background of SSP126 increased by 63.55% and decreased by 26.07%, respectively, compared with that under the current background. The low suitable area increased only by 3.05%, 4.81%, and 2.52%, respectively. In the 2090s, under the different backgrounds of SSP126, SSP245, and SSP585, the total suitable area of C. multiflorus also increased. The highly and medium suitable area of C. multiflorus under the background of SSP585 increased by 65.61% and decreased by 14.07%, respectively, compared with that under the current background. The low suitable area had no obvious change.

Under the SSP126 climate scenario, the total suitable area in the 2090s decreased by 2.74 percentage points compared to the 2050s. Specifically, the medium and low suitable area increased by 5.62% and 23.90%, respectively, while the highly suitable area decreased by 61.86%. Under the SSP245 climate scenario, there was little difference in the total and different grades of suitable area between the 2050s and 2090s. Under the SSP585 climate scenario, the total suitable area in the 2090s increased by 4.84 percentage points compared to the 2050s. Specifically, the highly and low suitable area increased by 56.50% and 3.87%, respectively, while the medium suitable area decreased by 16.01%.

Overall, the SSP126, SSP245, and SSP585 climate scenarios showed that the total suitable area and highly suitable area in the 2050s and 2090s were larger than the current condition (except for SSP245-2090s). The differences in the medium and low suitable area (165.467×10⁴ km²~174.771×10⁴ km²) between the 2050s and 2090s under different climate scenarios were not significant compared to that under the current condition (160.484×10⁴ km²), except for SSP126-2050s.

The map of spatial change of geographical distribution pattern ( Figure 9 ; Table 4 ) showed that the top two climate scenarios with the highest growth rate of potential suitable areas for C. multiflorus in the future were SSP126-2050s (11.53%) and SSP585-2090s (15.69%), other future climate scenarios ranged from 5.19% to 5.84%. It is worth noting that under the SSP126-2050s scenario, potential suitable areas of C. multiflorus would first appear in Zhejiang, Fujian, and Jiangxi. The top two climate scenarios with the highest loss rate of potential suitable areas were SSP245-2090s (3.31%) and SSP245-2050s (2.91%), but all loss areas were intermittent and small areas. Overall, the retention rates of potential suitable areas for C. multiflorus under different future climate scenarios ranged from 96.69% to 98.84%, showing an increasing trend (1.88~14.53%).

The importance value is an indicator that describes the significance of a plant species in a community. A higher importance value generally implies that the species holds a crucial position in the community, with a larger population size and wider distribution range compared to other populations. Niche breadth reflects the adaptability of a population within an ecosystem (Cai et al., 2021). From Table 5 , we can observe that C. multiflorus had the highest importance value (10.48) within the shrub community. However, its niche breadth, as measured by the Levins index, ranks third (B_L = 0.59), and fourth according to the Shannon index (B_S =3.16).

As shown in Table 6 , within C. multiflorus shrub community, 105 species pairs were formed by the 15 major shrub plant species. The niche overlap index (O_ik ) ranged from 0.09 to 0.79. Among these pairs, 63 pairs (60% of total pairs) had insignificant overlap (0<O_ik < 0.5) and 42 pairs (40% of total pairs) had significant overlap (O_ik > 0.5), suggesting that the ecological habits of these species pairs were not highly similar. 9 pairs with an O_ik value above 0.7, and most of these pairs consisted of species with relatively wide niches. Among them, C. multiflorus-Philadelphus incanus had the highest niche overlap (0.79). There were 16 pairs with an O_ik value less than 0.2, and these pairs had relatively narrow niche breadth, such as Ostryopsis davidiana - Viburnum opulus subsp. calvescens (0.09), O. davidiana - Quercus mongolica (0.09), and Syringa reticulata subsp. amurensis - Q. mongolica (0.09). Overall, the average O_ik value was 0.43, relatively fewer pairs with high overlap.

This study employed the ENMeval 2.0 version, which successfully solved the problems present in previous versions of the MaxEnt model, such as lacking optimal parameter selection (Morales et al., 2017), and overfitting (Merow et al., 2014; Elith and Graham, 2009), which led to inaccurate results. The ENMeval 2.0 version also helps researchers in resolving issues related to insufficient model performance and parameterization reporting, heavy reliance on AICc for model selection, and underutilization of spatial cross-validation (Kass et al., 2021).Therefore, this study employed the ENMeval 2.0 package to determine the optimal parameters for the MaxEnt model, ultimately identifying FC as LQPTH and RM as 2.5. Based on the optimized MaxEnt model, the simulated current distribution range of C. multiflorus was roughly consistent with the records in the Flora of China. Moreover, the AUC value (0.875) exceeded 0.8, indicating a high level of credibility for this simulation and providing a research basis for subsequent investigations (Araújo et al., 2005; Tang et al., 2021; Mahmoodi et al., 2022). The result is also useful to the collections of C. multiflorus germplasm resources as well as seed introduction.

Jackknife test ( Figure 5 ) revealed that the top three variables significantly influencing the distribution of C. multiflorus were Annual precipitation (Bio12), Min air temperature of the coldest month (Bio6), and Mean air temperature of the coldest quarter (Bio11), which aligned with the environmental factors affecting the distribution of Haloxylon, a genus with drought-resistance (Zhang et al., 2022). Among these, the contribution rate and the importance value of Annual precipitation (Bio12) were the highest, indicating that in arid and semi-arid regions, precipitation is more important than temperature. This is consistent with the findings of Qin et al. (2023), who observed that extreme drought in certain areas consumed soil moisture reserves, restricting plant survival and growth. Furthermore, when Annual precipitation (Bio12) exceeded 150 mm, C. multiflorus can survive, and when it exceeded 500 mm, the survival rate was 0.60 ( Figure 6 ). This further indicates that C. multiflorus exhibits strong tolerance to drought and may serve as a pioneer species for ecological restoration in arid and semi-arid areas.

Climate factors can not completely determine the geographical distribution of C. multiflorus. Soil conditions, human activities, and other factors also play important roles in shaping the distribution of plants at different spatial scales (Shishir et al., 2020; Xu W. B. et al., 2019). For instance, the seeds of C. multiflorus have hard outer shells, waxy inner membranes, poor permeability, and long physiological dormancy periods (Yu et al., 2017). Some scholars found that seeds not subjected to cold storage will not germinate the following year (Chi et al., 2015). These seed properties may influence the growth and distribution of C. multiflorus, consistent with the findings of Zhou et al. (2016) and Alzate et al. (2023). Additionally, it reproduces both sexually (inter-species hybridization) and asexually (Nybom and Bartish, 2007; Meng et al., 2021). This not only increases its genetic diversity and stress resistance to some extent but also effectively stabilizes hybrid advantages (Patel et al., 2018; Nitta et al., 2022), making it possible to adapt to various extreme climates. Therefore, in order to predict the distribution of plants accurately, it is necessary to consider more environmental factors and bio-ecological characteristics.

With the exponential growth of global greenhouse gas emissions, the trend of future climate warming in China will be further exacerbated (Li et al., 2020). C. multiflorus will be able to effectively respond to global warming. Under different future climate scenarios, the proportion of the total suitable area for C. multiflorus ranges from 43.64% to 48.98%, which is 0.8 to 6.14 percentage points higher than the current situation. This suggests that, within a certain range, moderate climate warming may facilitate species dispersal and population expansion (Bazzaz, 1990; Thomas et al., 2004; Schöb et al., 2009), especially benefiting drought-resistant shrub species (Screen, 2014).

Previous studies have shown that the climate scenario SSP245 is closest to the actual predicted climate conditions in China (Gao et al., 2013; Lai et al., 2023). However, under the climate scenario SSP126-2050s with the lowest carbon emissions, the total and highly suitable area of C. multiflorus increased by 4.06 and 11.53 percentage points compared to that of the current climate scenario, respectively. In addition to expanding to the high latitude areas of Liaoning in northeastern China and the southern part of Heilongjiang, C. multiflorus also expanded to the lower latitude regions of Anhui, and appeared for the first time in Zhejiang, Fujian, and some parts of Jiangxi. However, with the continuous increase in greenhouse gas emissions (SSP126-2090s), C. multiflorus disappeared in the above areas. This indicated that under low carbon emissions, these areas could meet the most suitable growth condition for C. multiflorus, but their advantages might disappeared when the climate turned warm.

Under all the climate scenarios, C. multiflorus expanded towards higher latitudes ( Figure 10 ), mainly towards the northern part of Xinjiang, Liaoning, and the southern part of Heilongjiang in northeastern China. This might be due to the fact that with the emission of greenhouse gases, the rainfall would gradually increase from the current altitude to higher altitudes (Jiang et al., 2022), while the temperature in temperate continental and cold temperate arid climate regions would also rise, which was more conducive to the growth of C. multiflorus. This is consistent with previous studies, which indicate that with the intensification of global warming, the distribution of most species will shift to higher latitudes and altitudes (Zhang et al., 2018; Faticov et al., 2021).

This study used MaxEnt model and CMIP6 data to simulate the potential suitable areas of C. multiflorus under different climate change scenarios. However, its distribution is not only affected by climatic factors, but also by population distribution (Theunis et al., 2005), community structure (Zhao et al., 2022) and many other factors. Therefore, when planning forest management and formulating restoration strategies, more attention should be paid to the differences of niche between species, which are crucial for explaining species coexistence at regional scales, as well as sustainable conservation and management of forests under climate change.

Niche analysis helps to understand coexistence mechanisms, predict community responses (Huang et al., 2019), and explain environmental tolerance of species (Smith et al., 2019). As shown in Tables 5 , 6 , the shrub community mainly consisted of 10 tree species from the Rosaceae, Betulaceae, Oleaceae, Hydrangeaceae, Adoxaceae, Celastraceae, Cornaceae, Caprifoliaceae, and Fagaceae families, indicating high species diversity and strong structural stability. C. multiflorus had the highest importance value (10.48), and could coexist with other tree species, demonstrating strong adaptability and dominance in community.

Interestingly, the presence of two woody tree species, Cornus macrophylla and Q. mongolica, within the shrub community suggested that the C. multiflorus community could adapt to relatively harsh conditions and begin to accumulate soil moisture and nutrients, promoting the construction of forest communities. The niche breadth of C. multiflorus ranked third, and the Shannon index ranked fourth with significant overlap with species such as Rosa bella, Syringa oblata, P. incanus, Euonymus phellomanus, Viburnum mongolicum, O. davidiana, and Spiraea ouensanensis, suggesting that these species may have a competitive relationship with C. multiflorus. Therefore, in ecologically fragile environments, it is advisable to first establish a C. multiflorus community, and avoid species with significantly overlapping niches, such as R. bella, S. oblata, P. incanus, E. phellomanus, V. mongolicum, O. davidiana, and S. ouensanensis.