Evaluating the Inﬂuence of Climate Change on Sophora moorcroftiana (Benth.) Baker Habitat Distribution on the Tibetan Plateau Using Maximum Entropy Model

: The ecosystems across the Tibetan Plateau are changing rapidly in response to climate change, which poses unprecedented challenges for the control and mitigation of desertiﬁcation on the Tibetan Plateau. Sophora moorcroftiana (Benth.) Baker is a drought-resistant plant species that has great potential to be used for desertiﬁcation and soil degradation control on the Tibetan Plateau. In this study, using a maximum entropy (MaxEnt) niche model, we characterized the habitat distribution of S. moorcroftiana on the Tibetan Plateau under both current and future climate scenarios. To construct a robust model, 242 population occurrence records, gathered from our ﬁeld surveys, historical data records, and a literature review, were used to calibrate the MaxEnt model. Our results showed that, under current environmental conditions, the habitat of S. moorcroftiana was concentrated in regions along the Yarlung Tsangpo, Lancang, and Jinsha rivers on the Tibetan Plateau. Elevation, isothermality, and minimal air temperature of the coldest month played a dominant role in determining the habitat distribution of S. moorcroftiana. Under future climate scenarios, the increased air temperature was likely to beneﬁt the expansion of S. moorcroftiana over the short term, but, in the long run, continued warming may restrict the growth of S. moorcroftiana and lead to a contraction in its habitat. Importantly, the Yarlung Tsangpo River valley was found to be the core habitat of S. moorcroftiana , and this habitat moved westwards along the Yarlung Tsangpo River under future climate scenarios, but did not detach from it. This ﬁnding suggests that, with the current pace of climate change, an increase in efforts to protect and cultivate S. moorcroftiana is necessary and critical to control desertiﬁcation on the Tibetan Plateau.


Introduction
Over the last century, anthropogenic warming has caused substantial changes in the spatial and temporal patterns of the global climate [1,2]. The resulting variations in extreme temperatures and precipitation, in particular, threaten the stability and diversity of plant ecosystems globally [3,4], especially in high-elevation regions [5]. As the world's third pole, the Tibetan Plateau has experienced more rapid climate change than other biomes on earth [5]. Previous studies have shown that the Himalayas have warmed, over the past 100 years, much higher than the global average of 0.74 • C [6]. The increase in the air temperature of the Tibetan Plateau also averaged 0.16 • C per decade from 1955 to 1996, with a particularly higher rate in winter (0.32 • C/decade), exceeding that of the areas at S. moorcroftiana habitat distribution; and (2) exploring potential range shifts of S. moorcroftiana habitat under future climate scenarios on the Tibetan Plateau. This study will improve our understanding of the processes and mechanisms of plant adaptation and dispersal under the complex climatic and environmental conditions in the plateau and provide a new theoretical basis and guidance for the use of S. moorcroftiana to control desertification on the Tibetan Plateau.

Study Area
Our study area included the entire Tibetan Plateau (26.00 • -39.78 • N, 73.31 • -104.78 • E; Figure 1) located in southwestern China. The Tibetan Plateau is approximately 2800 km in length from east to west and 300 to 1500 km in width from north to south, which comprises an area of~2.50 × 10 6 km 2 . The average elevation of the plateau is >4000 m. The climate of the Tibetan Plateau is characterized by generally strong radiation [42] and low air temperature, with decreasing air temperature as altitude increases, and, particularly, a large diurnal variation in ambient environments [43,44]. Both precipitation and temperature exhibit strong regional distribution patterns across the Tibetan Plateau, where annual precipitation increases from approximately 50 to 2000 mm from the northwest to the southeast, and annual air temperature increases from −15 to 20 • C from the northeast to the southeast [43]. the potential habitat distribution and shifts of S. moorcroftiana. There were two objectives for this study: (1) determining the impacts of bioclimatic, topographic, and soil variables on S. moorcroftiana habitat distribution; and (2) exploring potential range shifts of S. moorcroftiana habitat under future climate scenarios on the Tibetan Plateau. This study will improve our understanding of the processes and mechanisms of plant adaptation and dispersal under the complex climatic and environmental conditions in the plateau and provide a new theoretical basis and guidance for the use of S. moorcroftiana to control desertification on the Tibetan Plateau.

Study Area
Our study area included the entire Tibetan Plateau (26.00°-39.78° N, 73.31°-104.78° E; Figure 1) located in southwestern China. The Tibetan Plateau is approximately 2800 km in length from east to west and 300 to 1500 km in width from north to south, which comprises an area of ~2.50 × 10 6 km 2 . The average elevation of the plateau is >4000 m. The climate of the Tibetan Plateau is characterized by generally strong radiation [42] and low air temperature, with decreasing air temperature as altitude increases, and, particularly, a large diurnal variation in ambient environments [43,44]. Both precipitation and temperature exhibit strong regional distribution patterns across the Tibetan Plateau, where annual precipitation increases from approximately 50 to 2000 mm from the northwest to the southeast, and annual air temperature increases from −15 to 20 °C from the northeast to the southeast [43].

S. moorcroftiana Occurrence Records
In this study, we gathered a total of 242 effective population occurrence records of S. moorcroftiana across the Tibetan Plateau using four different data resources (Figure 1), including: (1) 36 from our field surveys in August 2020; (2) 37 from the Chinese National Plant Specimen Resource Center (CVH; http://www.cvh.ac.cn/, accessed on 13 October 2020); (3) 19 from the Forest Science Research Institute of Tibet Municipality [15]; and (4)

S. moorcroftiana Occurrence Records
In this study, we gathered a total of 242 effective population occurrence records of S. moorcroftiana across the Tibetan Plateau using four different data resources (Figure 1), including: (1) 36 from our field surveys in August 2020; (2) 37 from the Chinese National Plant Specimen Resource Center (CVH; http://www.cvh.ac.cn/, accessed on 13 October 2020); (3) 19 from the Forest Science Research Institute of Tibet Municipality [15]; and (4) 148 from an extensive literature review. Please see Table S1 for a more detailed description of these datasets.
There was a degree of spatial clustering existing in our S. moorcroftiana occurrence records, particularly in the Yarlung Tsangpo River (YTR) valley region ( Figure 1). This clustering can lead to model overfitting that exaggerates the model's performance but reduces its generability [45,46]. To overcome this issue, we employed the spatially rated occurrence data tool of SDMtoolbox 2.0 (Jason L. Brown, New York, NY, USA) [47] to eliminate spatial clusters. Here, we set the spatial interval to 10 km [48], which resulted in a subset of 100 population occurrence records that were later used in MaxEnt-based ENM.

Environmental Factors
A total of 27 environmental factors were included in the MaxEnt model to determine the ecological niche of S. moorcroftiana on the Tibetan Plateau (Table 1). These factors spanned a wide range of bioclimatic, topographic, and soil variables that have been shown to be important to the survival and migration of plants. The bioclimatic factors for the current climate baseline (average for 1970-2000) were extracted from the 2.5 min resolution historical climate database in WorldClim (https://www.worldclim.org/data/worldclim21. html, accessed on 9 May 2020). Topographical factors were obtained from the Harmonized World Soil Database v1.2 from the Food and Agriculture Organization of the United Nations (http://www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/, accessed on 9 May 2020). Lastly, soil factors were obtained from the Center for Sustainability and the Global Environment dataset (https://nelson.wisc.edu/sage/, accessed on 9 May 2020). Temperature seasonality Bio5 Max. temperature of warmest month Bio6 Min. temperature of coldest month Bio7 Temperature annual range Bio8 Mean temperature of wettest quarter Bio9 Mean temperature of driest quarter Bio10 Mean temperature of warmest quarter Bio11 Mean temperature of coldest quarter Bio12 Annual precipitation Bio13 Precipitation of wettest month Bio14 Precipitation of driest month Bio15 Precipitation seasonality Bio16 Precipitation of wettest quarter Bio17 Precipitation of driest quarter Bio18 Precipitation of warmest quarter Bio19 Precipitation Future climate scenarios were obtained from the shared socioeconomic pathways (SSPs) scenarios from the BCC-CSM2-MR global climate database of Coupled Model Intercomparison Projects 6 (CMIP6) (https://www.worldclim.org/, accessed on 9 May 2020) [49,50]. Data for four climate scenarios were accessed, including the SSP1-2.6 scenario (SSP126), the SSP2-4.5 scenario (SSP245), the SSP3-7.0 scenario (SSP370), and the SSP5-8.5 scenario (SSP585). These scenarios are defined by the amount of greenhouse gases that will be emitted in the years to come [49,50] and have been broadly used by the IPCC and other modeling efforts to project climate change in the next century. We resampled all environmental factors (including bioclimatic, topographic, and soil factors) to a 2.5 min resolution by using ArcMap.

MaxEnt Model Parameterization and Evaluation
The MaxEnt model utilizes the maximum entropy principle and applies five different feature constraints (i.e., linear, product, hinge, quadratic, and threshold) to calculate the probability distribution of a species' occurrence [51,52]. In this study, we used SDMtoolbox 2.0 [47] to carry out MaxEnt in ArcMap version 10.5 and set 25% of the occurrence data as testing data and 75% of the occurrence data as training data, based on 100 population occurrence data and 20 environmental factors. To determine the key environmental factors driving S. moorcroftiana habitat distribution, we performed a jackknife permutation to rank the environmental factors based on percent contribution. To evaluate the performance of the constructed MaxEnt modeling, we leveraged a receiver operating characteristic (ROC) curve analysis, where the area under the receiver operating curve (AUC) was used to indicate the accuracy of the model's prediction [26,53]. The larger the AUC is, the more accurate the model prediction is. Here, we defined model performance using three criteria: poor (AUC < 0.8), good (AUC 0.90-0.95), and excellent (AUC > 0.95) [54,55].
After the model calculations were completed, we visualized the habitat distribution of S. moorcroftiana on ArcMap. The continuous suitability score (SS; 0-1) was also derived from the MaxEnt model result. Based on the SS, we classified the landscapes on the Tibetan Plateau into four classes: unsuitable habitat (<0.25), low-suitability habitat (0.25-0.50), suitable habitat (0.50-0.75), and high-suitability habitat (>0.75).

Quantifying the Magnitude and Direction of S. moorcroftiana Habitat Shifts
In order to quantify the habitat shift of S. moorcroftiana under climate change, we applied the constructed MaxEnt model to future climate scenarios (i.e., SSP126, SSP245, SSP370, and SSP585) and compared the predicted habitats with those from current climate conditions using SDMtoolbox in ArcMap [47]. Thereafter, we divided the results into three categories: (1) expansion, (2) no change, and (3) contraction. In addition, to determine the overall spatial change in the habitat of S. moorcroftiana, we calculated the spatial centroids of S. moorcroftiana habitats, using the Centroid Changes command in SDMtoolbox, for both current and future climate scenarios. These centroids were then projected as vector arrows in ArcMap 10.5 to represent the magnitude and direction of habitat change for S. moorcroftiana [47].

Habitat Distribution and Key Environmental Factors Driving S. moorcroftiana Distribution under Current Environmental Conditions
Suitable distribution modeling for S. moorcroftiana performed well, with an AUC of 0.984 for both model testing and training (Supplementary Figure S1), indicating good performance of the constructed MaxEnt model in predicting the suitable habitats of S. moorcroftiana under current environments. We also found that the distribution of predicted suitable habitats under the current environments matched well with the identified S. moorcroftiana population occurrence records.
The predicted potential habitat for S. moorcroftiana had a total area of 15.29 × 10 4 km 2 , concentrated in regions along the Yarlung Tsangpo, Lancang, and Jinsha rivers in the Qinghai-Tibet Plateau (Figure 2). In particular, habitats classified as suitable or low suitability occupied a majority of the predicted habitats (suitable: 3.54 × 10 4 km 2 , 70.67% of the predicted habitat area; low suitability: 10.81 × 10 4 km 2 , 23.12% of the predicted habitat area). Conversely, habitats with high suitability covered an area of only 0.06 × 10 4 km 2 (0.06% of the predicted habitat area).

Potential Distribution of S. moorcroftiana under Future Climate Scenarios
The potential suitable habitat for S. moorcroftiana during the period 2020-2100 was predicted under four future climate scenarios. We compared these predictions with the current suitable habitat (Figure 2). The potential habitat areas of S. moorcroftiana exhibited a contraction trend during the 21st century under most climate scenarios (Figures 3 and 4), especially under the SSP370 and SSP585 scenarios, where significant contraction occurred in the upper reaches of the YTR in Rikaze, the Selincuo in Nagqu, and the Hengduan Mountains in Sichuan Province, China ( Figure 3). The area of habitat expansion for S. moorcroftiana ranged from 7.13 × 10 4 km 2 (SSP370-2061-2080) to 14.28 × 10 4 km 2 (SSP585-2061-2080), while the areas of contraction ranged from 6.75 × 10 4 km 2 (SSP585-2041-2060) to 13.45 × 10 4 km 2 (SSP126-2041-2060) (Figure 4).

Spatial Shift in the Habitat of S. moorcroftiana during the 21st Century
The vectors between the present and future habitat centroids indicated that the magnitude and direction of the range shift of S. moorcroftiana varied under different future climate scenarios ( Figure 5). Under the current climate scenario, the geographical centroid of the potential habitat for S. moorcroftiana was located at 91.56° E, 29.46° N, north of the Yarlung Tsangpo River in the Tibet Autonomous Region, China ( Figure 5). The species was predicted to shift its habitat to the west or the north under all climate scenarios. Specifically, under the SSP245, SSP370, and SSP585 scenarios, the centroid shifted westwards gradually, in a direction parallel to the Yarlung Tsangpo valley, with the farthest shift to

Spatial Shift in the Habitat of S. moorcroftiana during the 21st Century
The vectors between the present and future habitat centroids indicated that the magnitude and direction of the range shift of S. moorcroftiana varied under different future climate scenarios ( Figure 5). Under the current climate scenario, the geographical centroid of the potential habitat for S. moorcroftiana was located at 91.56 • E, 29.46 • N, north of the Yarlung Tsangpo River in the Tibet Autonomous Region, China ( Figure 5). The species was predicted to shift its habitat to the west or the north under all climate scenarios. Specifically, under the SSP245, SSP370, and SSP585 scenarios, the centroid shifted westwards gradually, in a direction parallel to the Yarlung Tsangpo valley, with the farthest shift to Lhasa River near Qushui County, Tibet Autonomous Region, China (located at 90.84 • E, 29.39 • N) under the SSP370 scenario.

Discussion
This study predicted the potentially suitable habitats for S. moorcroftiana on the Tibetan Plateau under current and future climate scenarios with mean AUCs of 0.984 and 0.985, respectively. Therefore, we believe that our model performance is robust and adequate for understanding the suitable habitat distribution of S. moorcroftiana. To the best of our knowledge, this is the first study to analyze the suitable habitat distribution of S. moorcroftiana for the present and future by using the MaxEnt model.

Relationship between S. moorcroftiana Habitat Suitability and Environmental Variables
In this study, we found that the habitat of S. moorcroftiana was narrowly concentrated along the Yarlung Tsangpo, Lancang, and Jinsha rivers across the Tibetan Plateau. The most critical environmental factor determining the habitat of S. moorcroftiana was elevation, followed by isothermality and the minimum temperature of the coldest month. This defines a significant difference between S. moorcroftiana and Sophora davidii (Franch.) [11]. We also found that S. moorcroftiana is suitable for regions with elevation ranging from 3400 to 4250 m, isothermality thresholds from 43 to 48, and minimum temperature of the coldest month thresholds from −19 to −9 °C (Table 2). This finding suggests that alpine environments are critical for the survival of S. moorcroftiana. Previous ecological studies have shown that precipitation and temperature are two of the most important factors determining global plant distribution and growth [56], especially in the arid and semi-arid regions of the Tibetan Plateau. In this study, we found that S. moorcroftiana had a high demand for low temperature and isothermality but not for precipitation, which can be explained by the strong absorptive root system of S. moorcroftiana [17]. This is consistent with the niche

Discussion
This study predicted the potentially suitable habitats for S. moorcroftiana on the Tibetan Plateau under current and future climate scenarios with mean AUCs of 0.984 and 0.985, respectively. Therefore, we believe that our model performance is robust and adequate for understanding the suitable habitat distribution of S. moorcroftiana. To the best of our knowledge, this is the first study to analyze the suitable habitat distribution of S. moorcroftiana for the present and future by using the MaxEnt model.

Relationship between S. moorcroftiana Habitat Suitability and Environmental Variables
In this study, we found that the habitat of S. moorcroftiana was narrowly concentrated along the Yarlung Tsangpo, Lancang, and Jinsha rivers across the Tibetan Plateau. The most critical environmental factor determining the habitat of S. moorcroftiana was elevation, followed by isothermality and the minimum temperature of the coldest month. This defines a significant difference between S. moorcroftiana and Sophora davidii (Franch.) [11]. We also found that S. moorcroftiana is suitable for regions with elevation ranging from 3400 to 4250 m, isothermality thresholds from 43 to 48, and minimum temperature of the coldest month thresholds from −19 to −9 • C ( Table 2). This finding suggests that alpine environments are critical for the survival of S. moorcroftiana. Previous ecological studies have shown that precipitation and temperature are two of the most important factors determining global plant distribution and growth [56], especially in the arid and semi-arid regions of the Tibetan Plateau. In this study, we found that S. moorcroftiana had a high demand for low temperature and isothermality but not for precipitation, which can be explained by the strong absorptive root system of S. moorcroftiana [17]. This is consistent with the niche of typical drought-resistant shrub species [57]. In fact, the absence of excessive demand for precipitation is one of the most important traits that have allowed S. moorcroftiana to become a favored drought-resistant silvicultural species for desertification control on the Tibetan Plateau.

Response of Suitable Habitat Distribution to Future Climate Change
Climate is the most critical ecological factor for plants on a large scale, and changes in plant distribution are the clearest and most direct responses to the climate [48]. In addition, soil, competition, and disturbance also play a vital role in plant growth and distribution across the Tibetan Plateau. Future climate change is likely to change the structure and function of terrestrial ecosystems, which, in turn, could result in changes in the distribution of plant species [20,58]. The Tibetan Plateau has been shown to be more vulnerable to climate change than most biomes on earth, due to its unique geographical characteristics and fragile ecosystems. For example, Liu et al. [59] revealed that climate change can significantly affect forest composition, distribution, and aboveground biomass in the subalpine forests of the eastern Tibetan Plateau. Ma et al. [60] found that the habitat suitability of Stipa purpurea (Poaceae) tended to significantly increase from the 1990s to the 2050s but then to decline from the 2050s to the 2070s. In this study, we quantified the habitat change of S. moorcroftiana on the Tibetan Plateau under four future climate scenarios and found that the habitat of S. moorcroftiana remained relatively stable under the SSP126 scenario, with a slight increase during the period 2020-2100. However, the total habitat of S. moorcroftiana contracted under the SSP370 and SSP585 scenarios ( Figure 3). Temporally, the habitat of S. moorcroftiana was likely to expand after 2060, following a contraction during the period 2020-2060, under the SSP126 and SSP245 scenarios. In contrast, its habitat consistently contracted throughout the period 2020-2100 under the SSP370 and SSP585 scenarios (Figure 4). Previous studies indicated that the Tibetan Plateau had been overall becoming warmer and wetter over the past decades, due to climate change [61]. Additionally, there was a simulation which showed that global warming promoted vegetation growth [60,[62][63][64]; warming temperatures positively affect the alpine meadow area of the Tibetan Plateau by reinforcing photosynthetic capacity and extending the growing season. These findings suggest that a consistent increase in temperature will have negative impacts on the distribution of S. moorcroftiana over the long term [65], though it can benefit in the short term. This is likely because the high demand for lower temperature (as discussed above) restricts the growth of S. moorcroftiana in warm climate conditions and eventually leads to a contraction of its habitat range.
The distribution on both sides of the Yarlung Tsangpo River valley is generally sparse, with very low cover and canopy height and mostly bare rock, in a fragile and degraded state, accompanied by drought-tolerant herbaceous plants, such as Artemisia wellbyi Hemsl. et Pears., and no tree layer [15,66]. Wind erosion is evident in the Yarlung Tsangpo River valley area, forming landscapes such as sands and sand dunes, with shrub layers dominated by S. moorcroftiana, Caragana spinifera Kom., and Hedysarum scoparium Fisch. [17,67]. In addition, we found that the geographical centroid of the habitat of S. moorcroftiana was located near the Yarlung Tsangpo River valley under current climate conditions ( Figure 5). This centroid was projected to shift westwards gradually along the Yarlung Tsangpo River valley under future climate scenarios. This suggests that the Yarlung Tsangpo River basin is likely to be the core habitat of S. moorcroftiana. Therefore, an increase in efforts to protect and cultivate S. moorcroftiana within the Yarlung Tsangpo River basin is necessary and critical in order to prevent desertification and land degradation on the Tibetan Plateau.

Uncertainties
In this study, we ultimately applied 100 population occurrence records to model the habitat distribution of S. moorcroftiana under both current and future climate scenarios. However, it is noted that the sample size was relatively small compared to the studied area (i.e., the Tibetan Plateau). Moreover, although a variety of bioclimatic, soil, and topographic factors were used in the MaxEnt modeling, the distribution of S. moorcroftiana could also be affected by other biotic or abiotic factors that are not included in this study, such as seed dispersal ability, species interactions, grazing, human activities, and land use. Therefore, the realized habitat distribution of S. moorcroftiana may be more fluctuating. However, introducing a greater number of factors into the model may lead to more collinearity problems and inflation of AUC values. Nevertheless, our model-predicted current habitat distribution of S. moorcroftiana matched well with field observations of S. moorcroftiana, highlighting the effectiveness of our constructed MaxEnt model. With that in mind, the results of this study could be reliable for understanding the processes and mechanisms of plant adaptation and dispersal on the plateau and provide a theoretical basis and guidance for desertification control on the Tibetan Plateau.

Conclusions
S. moorcroftiana is currently the preferred desertification control shrub species, due to its strong adaptability to sand burial and drought resistance on the plateau. To the best of our knowledge, this paper presents the first study on the distribution of S. moorcroftiana habitats on the Tibetan Plateau under both current and future climate change scenarios using ecological niche modeling. The habitat of S. moorcroftiana was found to be narrowly distributed along the Yarlung Tsangpo, Lancang, and Jinsha rivers on the Tibetan Plateau under the current climate. S. moorcroftiana is suitable for regions with elevation ranging from 3400 to 4250 m, isothermality thresholds from 43 to 48, and minimum temperature of the coldest month thresholds from −19 to −9 • C. In addition, warming temperatures can benefit the expansion of S. moorcroftiana in the short term but reduces its habitat range in the long term, likely due to its high demand for lower temperatures that restricts its adaption to continued warming. The Yarlung Tsangpo River valley was found to be the core habitat of S. moorcroftiana. This habitat will move westwards along the Yarlung Tsangpo River, but not detached from the Yarlung Tsangpo River valley, under future climate change. Therefore, an increase in efforts to protect and cultivate S. moorcroftiana within the Yarlung Tsangpo River basin is necessary and critical in order to prevent desertification and land degradation on the Tibetan Plateau.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/f12091230/s1, Figure S1: The ROC curve of S. moorcroftiana MaxEnt model under the current environment. Table S1: Geographical distribution and data sources of S. moorcroftiana population occurrences.