After the environmental variables were selected and the model simulation, the mean AUC values of H. r. sinensis and H. r. turkestanica were 0.927 (Figure 2A), and 0.969 (Figure 2B), and the AUC standard deviation were 0.004 and 0.009, respectively.

The contribution rate of each environmental variable is shown in Table 1. And they are average value over replicate runs. For H. r. sinensis, the top five environmental variables contributing to species distribution were precipitation of the wettest month (bio13, 28.83% contribution rate), mean temperature of the coldest quarter (bio11, 25.33%), isothermality (bio3, 18.54%), topsoil calcium carbonate fraction (t-caco3, 9.60%) and precipitation seasonality (bio15, 8.03%), their cumulative contribution was 90.33%. For H. r. turkestanica, the top six environmental variables contributing to species distribution were precipitation of the wettest month (bio13, 18.59% contribution rate), mean diurnal range (bio2, 13.49%), precipitation seasonality (bio15, 11.92%), topsoil cation exchange capacity of soil (t-cec-soil, 10.51%), topsoil calcium carbonate fraction (t-caco3, 10.31%), and topsoil PH (h2o) (t-ph-h2o, 7.22%), their cumulative contribution was 72.04%.

According to the response curves of the dominant environmental variables, for H. r. sinensis, we found that the highest habitat suitability was obtained when bio13 is 87–135 mm, bio 11 is −9.35 to −0.37°C; bio 3 is 28.3–31.9 and 41.5–45.3 (Figures 3A–C). When a location has a value above or below this value, its suitability decreases rapidly. And the relationship between habitat suitability and bio13 showed a positive skewness distribution, with the estimated peak value of 106 mm. For H. r. turkestanica, we found that the three key environmental variables affecting its distribution, which are bio13, bio2, and bio15, with values of 0–57 mm, 12.6–14.9°C, and 12.1–79.5, respectively (Figures 3D–F). Its habitat suitability is positively skewed with bio13 when values are higher or lower than the peak, which is 3.5 mm. We found that it has a narrow niche overall in bio13, its adaptability becomes poor beyond this range, and has a large adaptation range in bio15, and the probability of species growth decreases with increasing the value of bio15.

The distribution maps of the suitable habitat of H. r. sinensis and H. r. turkestanica under the current and future climatic scenarios are shown in Figures 4, 5, respectively. The results showed that the suitable area of H. r. sinensis is 883.19 × 10³ km² under current climatic scenario, accounts for 9.20% of China (Table 3), which mainly located in the Loess Plateau and the marginal area from southeast to northeast of the Qinghai-Tibet Plateau. A belt is formed from southwest to northeast, including most areas of Shanxi, central and northern Shaanxi, southwestern Sichuan, Ningxia, southern and Eastern Gansu, Northeastern Qinghai, central Inner Mongolia, and SouthEastern Tibet (Figure 4A). The low-suitability area is approximately 1444.01 × 10³ km², accounting for 15.04% of China. The suitable habitat area of H. r. turkestanica is 227.15 × 10³ km², accounting for 2.37% of China (Table 3), which mainly located around the Tarim Basin and Jungar Basin in Xinjing, western Ali, Shigatse and Qamdo in Tibet, Haixi Mongolian and Tibetan Autonomous Prefecture in Qinghai; Ganzi and Aba prefectures in Sichuan, and parts of Gansu and Ningxia (Figure 5A). The low-suitability area is approximately 1172.39 × 10³ km², accounting for 12.21% of the study area. The low suitable habitat is distributed at the periphery of the suitable habitat.

Under the different climatic scenarios in the 2050s, although the suitable habitat distribution of these two subspecies is similar to the current distribution (Figures 4B–E, 5B–E), they showed different increasing trends in the area of suitable habitat. For H. r. sinensis, the area of suitable habitat increased to 1033.76 × 10³, 1122.55 × 10³, 1129.54 × 10³, and 1116.97 × 10³ km² under the climatic scenarios of SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5, respectively. The average suitable habitat area is 1100.71 × 10³ km², accounting for 11.47% of the study area, an increase of 217.52 × 10³ km² from current. For H. r. turkestanica, the area of suitable habitat increased to 303.01 × 10³, 500.38 × 10³, 446.18 × 10³, and 410.29 × 10³ km², and the average suitable habitat area is 414.97 × 10³ km², accounting for 4.32% of the study area. Their unsuitable habitat areas also will be reduced (Table 3).

Under the different future climate scenarios, significant dynamic changes occur for both H. r. sinensis and H. r. turkestanica compared to the current climatic scenario. For H. r. sinensis, degraded suitable habitat is mainly distributed in Hebei, Shangdong and Henan. At the same time, the suitability of some areas in Eastern inner Mongolia will experience a decrease significantly under the SSP5-8.5. The increased suitable habitat is mainly distributed in Tibet, Qinghai, and Inner Mongolia, and with the increase of emission concentration, a belt area is formed in Inner Mongolia from southeast of Alxa left banner to southwest of Hailar (Figures 6A–D). The increased suitable habitat of H. r. turkestanica is mainly distributed around the Jungar Basin, Altay and Tacheng, the Tianshan Mountains, and Turpan Basin in Xinjiang, central Changdu in Tibet, Western and central of Inner Mongolia, Jiuquan in Gansu, southwest of Sichuan in the future. Under SSP2-4.5, the area of suitable habitat increased the most, and mainly in Xinjiang. These are the main loss areas of suitable habitat, such as Xifeng in Gansu, Yulin and Yanan in Shaanxi, and the southern of Shanxi (Figures 6E–H).

The habitats suitable for these two subspecies are relatively stable. The percentage of the stable suitable habitat for H. r. sinensis is slightly higher than that for H. r. turkestanica: The average value of the former is 79.06%, and the value of latter is 76.90% (Table 4). The results showed that the change in habitat area of the two subspecies from one species are slightly different under the different scenarios. The stability of suitable habitat for H. r. sinensis will show a slight decrease trend as the concentration of gas emissions increases. While H. r. turkestanica will be the least stable under SSP2-4.5, and the most stable under SSP1-2.6.

Figure 7 shows that the straight-line direction and distance of geographic migration of the suitable habitat centroids are heterogeneous. Currently, the centroid of suitable habitat is located in Guyuan county in southern Ningxia. While the centroids will gradually migrate westward with increasing emission concentrations under different future climate scenarios, and are distributed in Dingxi county, Linxia county and Xiahe county in Southeastern Gansu, respectively (Figure 7A). Overall, H. r. sinensis may migrate to the lower latitudes in the future. The centroids of H. r. turkestanica are located in Qiemo county, southeast Xinjiang. Currently, the centroid is located at 85°43′41″E, 38°21′26″N. Except for the centroid of the SSP2-4.5 scenario, the centroids will migrate to the higher latitude, the others migrate to the lower latitude (Figure 7B). The main migration direction will toward Eastern, and the magnitude of migration is greater under SSP2-4.5 and SSP3-7.0.

Although many methods can be used for understanding a species’ ecological niche distribution and for predicting the potential suitable habitat, the single model is still widely used. Among the SDM studies, MaxEnt is more effective in simulating the habitat suitability of plant species (Cobben et al., 2015; Ghareghan et al., 2020; Ab Lah et al., 2021). So, we chose this method for our study. The results showed that MaxEnt model can effectively simulate and predict the habitats suitable of H. r. sinensis and H. r. turkestanica. Therefore, we believed that the conclusions of this study are reliable. MaxEnt also was used to study the suitable distribution of H. r. sinensis by Li and Huang, who showed that the model achieved good simulation accuracy (Li et al., 2015; Huang et al., 2018). Despite MaxEnt having many advantages, some scholars think that a single model can produce over-fitting, and an integrated model may reduce the uncertainty of model fitting (Battini et al., 2019). The ensemble model (EM) strategy is widely used for studying the habitat distribution of species because of its higher accuracy, and better universality and fitting (Guo et al., 2019; Zhao et al., 2021). Moreover, MaxEnt is only applicable when considering abiotic factors; its applicability to areas strongly affected by human activities needs to be further verified. In future studies, other factors potentially affecting the distribution of species, such as human activities, biological interactions, diffusion constraints, and terrain variables (e g., landform classes, hill shade, heat load index, and compound topographic index) (Driver et al., 2020), should be more comprehensively considered. In this way, the results will be more accurate and the suitable habitat may be more fragmented.

Climatic elements have more influence on the growth and survival of these two subspecies of Hippophae rhamnoides than soil elements, especially H. r. sinensis. Bio13, bio11, and bio3 are the most important factors affecting the distribution of H. r. sinensis in this study. H. r. sinensis is affected by precipitation and temperature, and is less affected by soil, so it can be grown in meadow soil, chernozem, sandy soil and even in half-stone, sandstone, and sandy soil areas. The annual precipitation has a significant effect on its distribution: the 400 mm contour is the dividing line of the distribution area of H. r. sinensis on the Qinghai-Tibet Plateau. The Yunnan-Guizhou Plateau is unsuitable for its growth because its precipitation is too much, which is beyond the suitable range, so it does not extend to this area, which contradicts the research conclusion of Li et al. (2015). Whereas, Huang found that the precipitation of the coldest quarter has an important effect (Huang et al., 2018), it did not have significant effect in our study. This difference may be due to the different environmental variables selected and the future climate variables from different climate models. Precipitation and temperature are still the dominant factors affecting the geographical distribution of H. r. turkestanica. Its range of suitable habitat is narrower than that of H. r. sinensis, but its requirement for bio 13 is lower than that of H. r. sinensis, which may be the reason why the area of habitat suitable for H. r. turkestanica smaller than that of H. r. sinensis. H. r. turkestanica grows more easily under extreme drought conditions than H. r. sinensis.

Under the current climatic scenario, the habitat suitable for H. r. sinensis is concentrated in the middle and upper reaches of the Yellow River, the Loess Plateau, and the Northeastern to Southeastern margin of the Qinghai-Tibet Plateau, from the northeast of Hengduan Mountains to Taihang Mountains, north of Qinling Mountains and south of Yinshan Mountains, and it is distributed in a belt from Southwest to Northeast (Figure 4 and Table 3). The results are similar to that of Li’ results, but there are some differences in the area of suitable habitat, probably due to the number of selected sampling points and the suitability classification criteria applied (Li et al., 2015). H. r. sinensis likes sunlight and has certain water resources requirements, and distributes on banks of rivers, mountains, and valleys, and grows well in sparse forest land (Fischer et al., 2005). According to the difference in the climate and vegetation zonality in different growing areas of H. r. sinensis in China, four typical distribution areas can be identified: the Western Sichuan Plateau and Qinba Mountain distribution area, the Southern Loess Plateau distribution area, the central and northern Loess Plateau and Mu Us sandy land and arsenic sandstone distribution area, and the Qinghai Plateau distribution area (Dai et al., 2011). This subspecies can grow in the arid and barren Loess Plateau and arsenic sandstone areas due to its strong root germination and rich rhizobia, which can continuously improve the soil texture through nitrogen fixation. As a primitive group of Hippophae, Hippophae rhamnoides originated from the Qinghai-Tibet Plateau, some scholars have speculated that it may have spread from northeast to northwest and north China along the Hengduan Mountains. In the mid-1960s, H. r. sinensis invaded Mu Us sandy land through birds, and occupied the habitat with good water conditions in the area, and rapidly diffused to the slope and top of sand dunes, then formed a local unique ecological barrier. In recent years, H. r. sinensis has cultivated in Heilongjiang, Jilin, Shandong, Henan, and Beijing.

The suitable habitat area of H. r. turkestanica accounts for 2.37% of the total territory of China, and which is mainly distributed in Xinjiang, Tibet, and Gansu (Figure 5 and Table 3). A small amount of artificial cultivation occurs in Qinghai, Ningxia, and Inner Mongolia. It is often found in River Valley terraces, floodplains, and open hillsides. There are banded natural communities in the Tahe River Basin of Southern Xinjiang, along the Hotan River, bortara River Valley, both sides of the river in the Ali Region of Tibet, and Danghe river of Gansu. In addition, it shows drought resistance and barren resistance, and can grow in saline alkali beach and desert. For example, natural H. r. turkestanica grows in Taklimakan Desert. The two subspecies do not have high soil requirements, but some researchers showed that the growth of H. r. turkestanica is not as good as that of H. r. sinensis under the arsenic sandstone conditions.

Different regions have already experienced climate change and will continue in the future. Climate change significantly affects species survival. Species responses to climate change are heterogeneous under different climatic scenarios (Dyderski et al., 2018). Numerous studies have shown that climate change will not harm all species and the range of suitable habitat for most species will decrease and a few will expand. There is no general tendency for even closely related plants in the respond to climate change (Kolanowska et al., 2017). For the two subspecies considered in this study, the predicted range of suitable habitat under different climate scenarios in the future is basically consistent with the currently suitable habitat range, but shows significant dynamic changes. Their suitable habitat showed an increasing trend in the future climate scenario. But, the suitable habitat range, the direction and distance of centroid migration of these subspecies were heterogeneous (Figures 5–7). The increased suitable habitat of H. r. sinensis are mainly in Tibet, Qinghai, Gansu and Inner Mongolia, which are formed a belt in Inner Mongolia. The centroids of suitable habitat will move to the lower latitudes in the future and gradually migrate westward with increasing emission concentrations. H. r. turkestanica’s change also showed the same trend under the four climatic scenarios in the future, and its suitable range will increase. The centroids will move to the Eastern and lower latitude overall. With global warming, the area of arid and semi-arid areas will gradually increase (Huang et al., 2016). This may also be the reason for the expansion of suitable habitat of H. r. turkestanica in the future. Nevertheless, many studies had shown that species will move to places that are suitable for them, usually migrate to higher elevations and higher latitudes with climate warming (Li et al., 2013; Bezeng et al., 2017; Dyderski et al., 2018). The findings of this study do not support the view. The question of whether the response of subspecies to climate change can reflect the characteristics of the whole species needs to be further explored.

As a drought-tolerant plant with low requirement for natural environment, great adaptability, and with ecological, economic and social values, they are excellent species vigorously promoted for cultivation in China. In recent years, they have been successfully planted on a large scale on different types of degraded land, and the area of planted forests has been increasing year by year. However, extensive mortality has occurred in many artificial planting areas, mainly due to blind site selection and seed selection. For example, the central region of Inner Mongolia is only suitable for Hippophae rhamnoides subsp. mongolica, while Hippophae thibetana Schlechtend. and Hippophae gyantsensis (Rousi) Y.S. Lian cannot survive. Increased suitable habitat can be used as future artificial planting areas. The ecological niche of H. r. sinensis is wider than that of H. r. turkestanica, and more artificially planted, which has played a strong ecological role in the loess plateau, desert, watershed and other ecologically fragile zones and soil erosion areas. For most species, the suitable habitats are generally stable and will not face the risk of extinction with climate warming, exception of endangered species. Similarly, these two subspecies are not at risk of extinction in the future, and H. r. turkestanica is more vulnerable to climate change than H. r. sinensis. This is good news for the future planting of Hippophae rhamnoides Linn. in China. After 7–8 years of planting, Hippophae rhamnoides Linn. can form a plant community with more than 80% coverage, which greatly improves the local soil and water conservation capacity, increases soil fertility, improves biodiversity and stabilizes the ecosystem. The variation of suitable habitats of H. r. sinensis and H. r. turkestanica under different climatic scenarios can provide theoretical guidance and data support for the conservation of wild resources and the delineation of artificially expanded planting areas.