Effects of climate warming and human activities on the distribution patterns of Fritillaria unibracteata in eastern Qinghai-Tibetan Plateau

Fritillaria unibracteata is an endangered medicinal material species endemic to the Qinghai Tibet Plateau, and belongs to the national Class III endangered plant. In addition to expelling wind and removing damne, it also warms menstruation and relieves pain in clinic use of tranditional Chinese medicine. In recent years, affected by the destruction of shrubs and climate change, the habitat of F. unibracteata wild resources has been seriously damaged, indicating of great significance to predict its potential suitable habitat using MaxEnt model. The AUC values without human activities were 0.983 ± 0.013–0.988 ± 0.001, while it is 0.982 ± 0.015–0.989 ± 0.000 with human activities, justifying their applications for predicting the potential areas of F. unibracteata. Without human activities, there were 8.47 × 104 km2 of highly suitable habitats in northern Sichuan, southern Gansu and southeastern Qinghai. But the poorly, moderately and highly suitable areas of F. unibracteata have decreased to 33.8 × 104 km2, 9.66 × 104 km2 and 6.64 × 104 km2 due to human activities. Environmental variables affecting F. unibracteata distribution included the minimum temperature in the coldest month (−16.89–−4.96 °C), annual precipitation (416.64–866.96 mm), temperature annual range (24.83–31.97 °C), elevation (2879.69–3981.82 m), human footprint (2.58–23.66) and mean UV-B of highest month (7381.92–8574.27 kJ/m2). In the 2050s and 2090s, human activities would significantly reduce the highly suitable habitats of F. unibracteata. Under SSP1-2.6, SSP2-4.5 and SSP5-8.5 scenarios, the centroid would move to the low latitude area from the current position first, and then to a high latitude area. Wild resources of F. unibracteata in China can be effectively conserved based on our results.

rostratum 9 .Different human activity ranges, ways, and intensities can lead to different types of land use (arable land, forest land, shrubland land, and residential land), which directly affects the spatial distribution and diversity of plants 10,11 .Therefore, taking human activities as prediction variables can more accurately show the potential distribution range and spatial pattern of species in the future.
MaxEnt (the maximum entropy model) invented by Jaynes in 1957, can infer incomplete information, and was used to predict species distributions in 2004.In today's world, MaxEnt has been widely applied in a wide range of fields, including ecology, evolution, and resource management 12,13 .MaxEnt was more accurate than other niche models, including CLIMEX, GARP, Bioclim and Domain.In spite of incomplete data on species distribution and environmental variables, it can accurately predict distribution areas of species 14 .Additionally, MaxEnt has high stability and is applied in the field of animal and plant conservation, including predicting potential habitats for medicinal plants and endangered species 15,16 .She et al. predicted the distribution hot spots of Notopterygium incisum, an important medicinal resource and endangered species in the Three Rivers Source Area under the future climate change scenarios using MaxEnt 17 .Ji et al. 18 applied MaxEnt and combined with statistical methods to simulate the current and future distribution of Paris verticillata, explored the relationship between its geographical distribution, and understood the development trend of its geographical distribution in the future.
Fritillaria unibracteata P.K.Hsiao & K.C.Hsia (Liliaceae: Fritillaria) is a perennial herb (Fig. 1).The dried bulbs of F. unibracteata with important medicinal value are widely used to treat lung heat and dry cough, dry cough with less phlegm, yin deficiency and labor cough, expectoration with blood 19 .In addition, it can also reduce the permeability of blood vessels and effectively achieve the anti-inflammatory effect 19,20 .The total saponins and total alkaloids in bulbs have a certain antihypertensive effect on the cardiovascular system 19,21 .There is a growing market demand for F. unibracteata since it has better efficacy than other basic species of Chuanbei, such as Fritillaria przewalskii, Fritillaria cirrhosa and Fritillaria delavayi.As a medicinal plant, the previous research on F. unibracteata not only involves its biological characteristics, chemical components and pharmacological effects, but also involves its community ecology.Huang et al. 22 found that the distribution of F. unibracteata is in the transition from China-Japan forest plant subregion to China-Himalayan plant subregion, which is the most abundant area of alpine flora in the world.Due to the narrow niche and long growth cycle, as well as the deterioration of the ecological environment caused by climate change and people's excessive excavation, the wild resources of F. unibracteata have been greatly reduced in recent years and have been listed as a national third-class endangered medicinal material species 22 .Currently, studies using models to simulate the habitat of F. unibracteata have not included human activities as an important variable.Wang et al. analyzed the suitable habitat of F. unibracteata in China by using TCMGIS-I system 23 .Two reports have applied MaxEnt model to simulate the distribution of F. unibracteata with the impact of current climate condition 24,25 .On the basis of www.nature.com/scientificreports/previous studies, we introduced human activities as a comparative variable, and our purpose was to solve the following issue: (1) analyze the key environmental variables affecting the potential distribution of F. unibracteata.
(2) simulate the distribution pattern of F. unibracteata in eastern Qinghai-Tibetan Plateau.(3) clarify the role of human activities in the geographical distribution of F. unibracteata.The results can provide data support for the wild cultivation and introduction of F. unibracteata, and provide theoretical basis for the protection and scientific research of the Fritillaria genus.

Importance of environmental variables
Without human activities, the percent contribution rate (44.54%) and the regularized training gain (1.96) of elevation (El) were the highest (Fig. 2A and Table 1), indicating that it was the most important variable for the distribution of F. unibracteata.When modelling with only variable, the gain (1.63) of min temperature of coldest month (Bio6) ranked second (Fig. 2A), while modelling without Bio6, the score of the model decreased the ) and annual precipitation (Bio12) exceed 1.3 (Fig. 2A), and their percent contribution rates reached 18.17% and 28.46% respectively (Table 1).With human activities, the percent contribution rate (45.63%) and the regularized training gain (1.95) of elevation (El) were the highest (Fig. 2B and Table 1), indicating its importance.The regularized training gain (1.58) and the percent contribution rate (15.04%) of mean UV-B of highest month (UV-B3) ranked second and third respectively (Fig. 2B and Table 1).The training gain value for "without variables" was most affected by mean UV-B of the highest month (UV-B3) (Fig. 2B), indicating that its unique information and great impact to the distribution of F. unibracteata.The regularized training gain of human footprint index was 1.48 (Fig. 2B), which showed the necessity of introducing it.
By the above comparison method, min temperature of coldest month (Bio6), elevation (El), annual precipitation (Bio12), human footprint index (Hf), mean UV-B of highest month (UV-B3) and temperature annual range (Bio7) were identified to be the dominant environmental variables.

Suitable value range of main environmental variables
When the elevation (El) < 3568.52 m, it had a positive impact on the presence probability of F. unibracteata, that was, the presence probability increased with elevation, while when the elevation > 3568.52 m, it had a negative impact on the presence probability (Fig. 3).When the min temperature of coldest month (Bio6) was −10.5 °C, the presence probability of F. unibracteata decreased steadily after reaching the peak (Fig. 3).The presence probability of F. unibracteata increased rapidly to the highest value 10.52 (P = 0.7) when the human footprint value reached 2.47, then followed by a decrease in probability.According to Fig. 3, the suitable ranges of min temperature of coldest month, annual precipitation, temperature annual range, elevation, human footprint and mean UV-B of highest month were −16.89-−4.96℃ (Fig. 3A), 416.64-866.96mm (Fig. 3B), 24.83-31.97°C (Fig. 3C), 2879.69-3981.82m (Fig. 3D), 2.58-23.66(Fig. 3E) and 7381.92-8574.27kJ/m 2 (Fig. 3F), respectively.
Under the influence of human activities, highly, moderately and poorly suitability areas of F. unibracteata were 6.64 × 10 4 km 2 , 9.66 × 10 4 km 2 and 33.8 × 10 4 km 2 , respectively (Fig. 4B and Table 2).The fragmentation of suitable habitats of F. unibracteata were more apparent under human activities.The areas of highly, moderately, and poorly suitable habitats decreased by 21.63%, 48.91%, and 30.84%, suggesting a negative correlation between the F. unibracteata's distribution and human activities.GAP analysis showed that the highly suitable habitats of F. unibracteata overlapping with the nature reserve was 2.34 × 10 4 km 2 without human activities (Fig. 4C).With human activities (Fig. 4D), the overlapping area was 1.84 × 10 4 km 2 , accounting for 27.71% of the highly suitable habitats.

Centroid variations of the suitable habitats under climate change scenarios
Without human activities (Fig. 9A), under SSP1-2.6, the centroid of the highly suitable habitats would move  43°E).By 2090s, the centroid would generally displace 14.4 km to the northwest (Fig. 9A).
With human activities (Fig. 9B), under SSP1-2.6, the centroid of the highly suitable habitats would displace 23.

Evaluation of models
Without human activities, based on the training data and test data of the MaxEnt under the current situation, the AUC values were 0.987 ± 0.001 and 0.983 ± 0.013 respectively.Under future climate scenarios, the AUC values for training data were 0.986 ± 0.001-0.988± 0.001, while for the test data were 0.979 ± 0.021-0.984± 0.006 (Table 3).
With human activities, the AUC values of the training data and test data of the MaxEnt model under current conditions were 0.987 ± 0.001 and 0.982 ± 0.015 respectively.Under climate change scenarios, the AUC values of the training data were 0.986 ± 0.001-0.989± 0.000, and that of the test data were 0.980 ± 0.015-0.984± 0.006 (Table 3).

Discussion
As a comprehensive environmental factor that affects plant growth and distribution, altitude largely controls the changes in other environmental factors, such as temperature, precipitation, and light intensity, which exhibit regular changes with altitude [26][27][28][29] .Altitude is an important factor affecting Fritillaria plants.Jiang et al. found that the dominant factors related to the growth of wild plants of Fritillaria cirrhosa are altitude, precipitation in September, precipitation in November, and vegetation type.Zhao et al. demonstrated through simulation that that although the dominant factors affecting the habitat suitability of F. cirrhosa, F. unibracteata, Fritillaria przewalskii, and Fritillaria delavayi are different, but altitude plays an important role in all species 24 .Fritillaria unibracteata is a kind of alpine plant, which mainly grows in the high altitude area of 2800-4400 m on the eastern edge of Qinghai Tibet Plateau.In this area, altitude is an important factor controlling the combination form and variation degree of other environmental and biological factors 30,31 .The growth status and survival strategy of plants growing in extreme alpine environment can show adaptability to altitude 32 .By comparing, the percent contribution rate of elevation was 44.54% without human activities and 45.63% with human activities, which were the most important variables under both scenarios.F. unibracteata grows and reproduces poorly at low altitudes, while the bulb biomass and total plant biomass decreased as altitude increases in high altitudes 33 .Xu et al. demonstrated that the effects of altitude gradient and life history stage on single leaf area, plant height, and specific leaf area of F. unibracteata were very significant, and there was obvious interaction 34 .Chen et al. believed that the reproductive organs of F. unibracteata will be affected by altitude, which is mainly related to snow melting time, temperature, ultraviolet intensity, flower visiting insects 35 .Therefore, altitude has a great restrictive effect on the growth of F. unibracteata.A response curve revealed that the suitable elevation range was 2879.7-3981.82m, confirming that altitude plays an important role in F. unibracteata distribution.Previous studies have shown  According to the statistical data, in the high-altitude areas of western Sichuan, southeast Qinghai and southern Gansu, the annual precipitation was about 400-1400 mm, and the temperature annual range was large, which was conducive to the accumulation of nutrients and promote the growth of bulb of F. unibracteata.Based on the response curve, the suitable range for the min temperature of coldest month was −16.84-4.95°C, while the average temperature during the coldest month was about −2 °C, which could ensure its safe wintering.For the northwest regions with relatively less rainfall (Xinjiang, Inner Mongolia and western Gansu), or the southern coastal areas with more rainfall, it is not conducive to the accumulation of nutrients of F. unibracteata, so its distribution in these areas is limited.In this study, the suitable range of main variables calculated by MaxEnt was basically consistent with the ecological environment distributed in the field, that was, high altitude, low temperature, thin air, strong sunshine, clear dry and rainy seasons and large temperature diurnal range 34,35 .
The distribution of species can be limited by climate factors at large scales.However, many studies have proved that the use of climate factors separately will lead to a wider predicted niche.Now the most viewpoints are that even in the suitable area, the high fragmentation of natural habitat caused by human activities will hinder the expansion of species.A study by Xu et al. assessed whether and how human activities have changed the extent of 9701 vascular plants' climatic potential ranges in China.and they found narrow-ranged species showed negative range-filling relationships to these human indicators 36 .Zhang et al. 's simulation showed that the modern suitable area of Rosa persica was significantly reduced after the addition of human footprint data 37 .Cao et al. compared the impact of human footprint on the suitability of Swertia przewalskii, a unique plant in the Qinghai Tibet Plateau, and revealed that after the introduction of this variable, the potential habitat of S. przewalskii species showed a fragmentation trend and the area decreased by 32% 6 .In view of this, in order to more comprehensively reflect the ecological environment of F. unibracteata and obtain accurate results, we not only selected the commonly used bioclimatic variables, but also supplemented altitude, soil factors and human activities.According to our results, human activities had a high percentage impact on the growth suitability of F. unibracteata (3.61%).It can be concluded that F. unibracteata's growth, habitat and reproduction may be affected by a series of human activities such as population density, land use, roads and railways as well as construction.
Under current condition, the highly suitable habitats of F. unibracteata were mostly located in northern Sichuan, southern Gansu and southeastern Qinghai and the total area was 6.64 6.64 × 10 4 km 2 .Field investigation and literature review showed that there were cultivation bases of F. unibracteata in Western Sichuan Plateau and southern Qinghai located in the predicted highly suitable habitats 38,39 .Liu et al. 's prediction suggested that the most suitable growth areas of F. unibracteata with an area of 5.62 × 10 4 km 2 were dominantly distributed in western and northern Sichuan, south Gansu and south Qinghai, which is consistent with our result 25 .Wang et al. predicted that the best suitable areas for F. unibracteata were Aba, Ganzi in Sichuan, and southwest Tibet, while Qinghai and Gansu were unsuitable areas, which was different from our results 23 .Our simulation showed that under the impact of human activities, the area of highly, moderately and poorly suitable habitats decreased by 21.63%, 48.91% and 30.84% respectively.Most people believed that even in areas with suitable climate, the high fragmentation of natural habitats caused by human over use of land would hinder the expansion of species habitats 5,40 .Combined with the prediction results and field observation, we speculated that an important reason for the area reduction of F. unibracteata was the destruction of land and habitat change caused by local urban expansion and road construction.In addition, animal husbandry was one of the important industries in Qinghai Tibet Plateau, therefore the impact of grazing could not be ignored.
Recently, it was found that the suitable areas of several medicinal plants were decreasing under future climate change scenarios.Ji et al. showed that under RCP 2.6, RCP 4.5, RCP 6.0 and RCP 8.5 scenarios, the total suitable area of Paris verticillata in China showed a decreasing trend in 2050s and 2070s 18 .A study by Liu et al. calculated the change in the highly suitable area of Fritillariae Cirrhosae Bulbus under SSP1-2.6,SSP2-4.5 and SSP5-8.5, and found that the highly suitable area decreased in size, and the northwest area would be the geometric center of the total suitable area 41 .Wei et al. confirmed that the most suitable area of Fritillaria walujewii will decrease in the middle and south of Yili River Valley and Tacheng area 42 .However, some scholars have reached different conclusions.Guan et al. believed that climate warming can promote the growth of some plants, and the suitable area of Quercus acutissima will increase in general under future climate conditions 43 .This study found that under human influences, F. unibracteata habitats are likely to decrease significantly in 2050s and 2090s.The size of species distribution areas is an important characteristic of species, and narrow range species often have specific environmental needs and may be more sensitive to habitat changes.Under climate change and intense human activities, the distribution area of narrow range species is likely to shrink, while the distribution area of widely distributed species is likely to expand [44][45][46][47] .F. unibracteata is a typical narrow niche plant, this may explain the significant reduction in its habitat after adding human activities index as environmental variable.
Under the three scenarios, the centroid of F. unibracteata would move from the current position to the low latitude area first, and then to the high latitude area.As a typical alpine plant, F. unibracteata has the characteristics of cold tolerance, humidity preference, fear of high temperature and shade preference 19,48 .We speculated that the change of suitable habitat and the moving trend of centroid in the future were the result of the dual influence of temperature and precipitation.Studies have shown that the increase of regional temperature and precipitation will destroy the habitat of some alpine plants 49,50 .By the 2090s, the increasing trend of temperature and precipitation in the Qinghai Tibet Plateau will be more obvious, and the suitability of the original growth area of F. unibracteata will be further reduced.
With the development of green and healthy industries, medicinal plant resources have attracted increasing attention.However, the contradiction between the demand for traditional Chinese medicine and the increasing scarcity of resources is becoming increasingly prominent 51,52 .Due to strict habitat requirements and indiscriminate mining, the wild resources of F. unibracteata have been depleted.Therefore, determining the optimal area www.nature.com/scientificreports/for artificial introduction and cultivation of medicinal plants through habitat suitability research is one of the effective methods to fundamentally alleviate the pressure on wild medicinal plant resources.The protection of rare and endangered plant resources is not only the key component of biodiversity protection and sustainable utilization, but also one of the important tasks of nature protection.In recent years, people have gradually realized that the loss of species habitat has become the biggest threat to biodiversity.The destruction of the original habitat by reservoirs, ponds and farmland has resulted in the disappearance of many wild populations of F. unibracteata in Qinghai and Sichuan 24,53,54 .F. unibracteata is urgently in need of better protection.Our GAP analysis showed that there is a large protection GAP of F. unibracteata in northern Sichuan, southeastern Qinghai and southwestern Gansu.GAP analysis also showed that Taohe, Siguniangshan, Wolong, Sanjiangyuan, Zoige and other more than ten national nature reserves were located within the highly suitable habitats.Therefore, combined with the distribution characteristics in National Nature Reserves and the current national reserve management policy, we put forward the following suggestions: (1) In the areas where the highly suitable habitats overlaps with the National Nature Reserve, the core protection area of F. unibracteata will be established to protect its habitat completely, and the core protection area will also play a role in the conservation and radiation of its germplasm resources.( 2) Studies revealed that it was feasible to transplant F. unibracteata artificially to a suitable habitat 48,55 .Therefore, for the moderately and poorly suitable habitats, in the areas with dense population, frequent human activities, which have caused serious harm to the groups and germplasm resources of F. unibracteata.Depending on the environmental conditions such as altitude, soil, temperature and precipitation, they can be transplanted into reservoirs, farms, and parks suitable for their survival and free of human disturbance so as to achieve the purpose of ex-situ genetic resources protection.(3) In the general control area of the National Nature Reserve, the field distribution survey of F. unibracteata should be further carried out, and the community survey should be carried out in relevant villages and towns to understand the use of local residents' natural resources in the community.On this basis, local residents are encouraged to understand, support and participate in the protection of F. unibracteata by carrying out environmental education, science dissemination and seeking alternative livelihoods.

Study area and occurrence records of F. unibracteata in China
The study area is predominantly in the eastern edge of the Qinghai Tibet Plateau (Fig. 10), with a large space span, belonging to the plateau marine climate and continental cold dry climate.The air in this area is thin, the annual average temperature is mostly 7 ~ 10 °C, the highest temperature can reach 30 ℃, and the lowest temperature is as low as −20 ~−30 °C.Seasonal frozen layer is distributed in the area above 4000 m, with the maximum thickness of 50 cm.The vegetation is mainly alpine meadow and the overall vegetation coverage is generally low.There is almost no vegetation in the area above 4500 m, with obvious vertical zoning 56 .Occurrence data of F. unibracteata were were acquired from the National Specimen Information Infrastructure (NSII, http:// www.nsii.org.cn/), the Chinese Virtual Herbarium (CBV, https:// www.cvh.ac.cn/), the Global Biodiversity Information Facility (GBIF, https:// www.gbif.org/), field survey (Our field survey only recorded the coordinates of occurrence data and did not involve field sampling), and literature.F. unibracteata distribution records were processed according to a literature review (Wang et al. 2019; Liu et al. 2021).The first step included using Baidu's coordinate picking system (https:// api.map.baidu.com/ lbsapi/ getpo int) to determine longitudes and latitudes accurate to the town level.The second step involved using Microsoft Excel (2010) to remove duplicate records.Third, each point was measured with respect to its cell grid center, and the nearest point was selected.To establish MaxEnt, 123 distribution points of F. unibracteata were retained after the above procedures (Fig. 10).

Environmental variables
Five independent environmental variables were included in this study (Table S1).Eliminating the influence of multicollinearity was important for simulation, so Pearson correlation coefficient was adopted 57,58 .We first calculated the percent contribution rate of 19 variables using MaxEnt, and we retained the variables with higher percent contribution rates (Table S2).Following that, SPSS was used to analyze Pearson's coefficients between two variables with percent contributions greater than 0 for 123 F. unibracteata occurrences.Thirdly, the higher coefficient was retained if the percentage contribution of the variables with the absolute value of the coefficient exceeded 0.85 (Table S3).The prediction model for F. unibracteata included 15 variables in addition to elevation (Table 4).

Modelling process
MaxEnt software operation procedure are presented as: (1) Occurrence data of F. unibracteata were imported into the "sample" and "environmental layer" data boxes of MaxEnt software (V3.4.4) in CSV and ASC formats (including bioclimatic variables, soil data, elevation and human footprint).(2) "Create response curves" and "Do jackknife to measure variable importance" were selected respectively to analyze the relationship between variables and presence probability of F. unibracteata and measure the importance of variables.(3) In the initial model, "Random test percentage" was set to 25%, while in the reconstructed model, "random seed" was selected, and the "replicates" was set to 10 12,59 .An area value under the ROC curve, such as AUC value ranges from 0.5 to 1, is a highly recognized diagnostic test evaluation index 60,61 .

Figure 2 .
Figure 2. Import of environmental variables for prediction based on jackknife test.

Figure 9 .
Figure 9. Variations of the centroids of the suitable habitats under climate change scenarios without human activities (A) and with human activities (B).ArcGIS v10.0: https:// www.arcgis.com/.

Table 1 .
Percent contribution of 15 environmental variables.

Table 3 .
AUC values of models.

Table 4 .
Environmental variables used to predict the potential geographic distribution of F. unibracteata.