Distribution Range and Richness of Plant Species Are Predicted to Increase by 2100 due to a Warmer and Wetter Climate in Northern China

ABSTRACT

The warming global climate is threatening terrestrial ecosystem stability, including plant community structure and diversity. However, it remains unclear how distribution, richness, and turnover of plant species are impacted by warming and wetting in northern China. In the present study, species distribution models were applied to predict the spatial distribution of 5111 plant species based on 111,071 occurrence records in northern China. Additionally, variations in species richness and turnover rates were predicted for 2100 under 3 scenarios. The results indicated that approximately 70% of plant species will expand in their distribution, resulting in an increase in species richness. These changes will be driven mainly by temperature seasonality (TSN), annual precipitation (MAP), and mean temperature of the coldest quarter (MTCQ). However, about 30%–40% of the species will face extinction risks, including a considerable number of endemic and Red‐Listed species, and suitable habitat loss (LSH) will exceed 30%. Narrow‐ranging species will be more likely to lose a larger percentage of their suitable habitats than wide‐ranging species, highlighting their sensitivity to environmental changes. Importantly, it emerged that species turnover rates will increase linearly with ecological vulnerability at the grid level, indicating that community structure and species composition are easily affected by climate change in ecologically vulnerable areas. Therefore, biodiversity hotspots with high species richness in the southern study areas, as well as regions exhibiting both fast species turnover and significant ecological vulnerability, should be prioritized for conservation. These findings provide insights into how species composition and richness in plant communities vary with global climate change and provide effective ecological conservation and management strategies.

Climate change is reshaping plant species distribution and richness. Based on data from 5111 plant species in northern China, our study predicts that many species will expand their range by 2100 due to warmer and wetter conditions, increasing species richness. However, about one‐third of species may lose much of their suitable habitat, including some narrow‐ranging or Red‐Listed species. Regions with high ecological vulnerability are expected to experience faster changes in species composition. These findings highlight the need to prioritize conservation in biodiversity hotspots and vulnerable regions, offering valuable guidance for managing ecosystems under future climate change.

1. Introduction

Climate and land cover changes are recognized as principal drivers of biodiversity shifts (Zhang et al. 2017; Song et al. 2021; Peng et al. 2021). Pressures of climate and land cover change not only compel species to migrate to higher elevations and latitudes (Chen et al. 2011; Hartikainen and Robson 2022), but also escalate biodiversity loss and the risk of species extinction (Powers and Jetz 2019; Yu et al. 2021; Holzmann et al. 2023). For example, altering temperature and precipitation patterns led to changes in the spatial distribution ranges of plant species, thereby affecting biodiversity. Changes in land cover can cause habitat loss, potentially exacerbating the impacts of climate change on biodiversity (Newbold et al. 2016). Northern China is characterized by vast arid and cold regions, where the unique natural environment has given rise to numerous endemic species with tolerance to stress factors such as salinity, extreme cold, and drought. These arid ecosystems also have substantial carbon storage capacity, making an important contribution to mitigating climate change (Dong, Ran, et al. 2024). However, the ecosystems in this region are sensitive to climate change (Huang et al. 2017).

Responses to climatic and land cover changes differ among plant lifeforms, reflecting their unique adaptations to environmental stressors. Within these lifeforms, narrow‐ and wide‐ranging species exhibit different ecological strategies. Narrow‐ranging species are more specialized and are confined to specific habitats (Slatyer et al. 2013), whereas wide‐ranging species have broader ecological niches, enabling them to thrive across diverse environments. Wu et al. (2023) analyzed time‐series data from 238 metacommunities to compare how the presence of narrow‐ and wide‐ranging species changed over time. They concluded that narrow‐ranging species reduced their range, while wide‐ranging species tended to expand their range. Understanding the future potential changes in distribution ranges and species richness of both narrow‐ and wide‐ranging species across various lifeforms would benefit biodiversity conservation efforts. Differences in ecological strategies of plant species and their responses to environmental changes are not only evident at the species level but are also reflected in patterns of plant richness and ecosystem functions. For example, the proportion of woody plants decreases while that of herbaceous plants increases when moving from the tropics to the poles (Taylor et al. 2023). Moreover, species with narrow ranges tend to inhabit more extreme environments than their wide‐ranging counterparts (Wu et al. 2023). For instance, deserts and cold regions are typically home to narrow‐ranging species, the former adapted to dry climates and the latter to cold climates (Yao et al. 2021, 2024).

Species distribution models are used widely to assess the spatial and temporal dynamics of species under global change (Elith and Leathwick 2009; Pillet et al. 2022). These models play a pivotal role in identifying biodiversity hotspots and predicting their future trends, as well as assessing extinction risks for endemic, endangered, and invasive species (Gallardo et al. 2017; Zhang et al. 2017; Guan et al. 2020; He et al. 2021; Yang et al. 2023). Numerous studies predicted spatiotemporal distributions and extinction risks of certain families, genera, or woody plants based on species distribution models (Peng et al. 2021; Xie et al. 2022; Zhao et al. 2023). Qiu et al. (2024) predicted the potential distributions of 32,000 species of plants in China using the Maxent model and developed an online platform for predicting plant distributions in China based on visualization technology. In a previous study, we employed random forest models to determine the spatiotemporal patterns of species richness for annual herbs, perennial herbs, and woody plants across China's drylands, from the last interglacial period to 2070, based on relationships between species richness, climatic variables, soil features, and human activities (Sun et al. 2021). However, the spatiotemporal distribution patterns of each plant species are still unknown. Importantly, we did not identify: (1) the variations in species composition and turnover rate in plant communities across drylands, and, thus, which species are expanding or contracting; and (2) richness hotspots and conservation priority areas under future climates. Species turnover reflects the dynamic changes in species composition across temporal or spatial scales. A faster turnover rate may indicate greater adaptability of communities, but it could also suggest a potential decline in ecosystem stability and a risk of impaired ecological functions. Addressing these gaps is crucial, as it offers key insights for targeted conservation strategies in the face of climate change.

Identifying priority areas and species for conservation is widely regarded as essential information to halt the decline in biodiversity. Conservation priority areas are either rich in biodiversity and not included in existing protected zones, or are at high risk of losing species (Chen et al. 2017). In addition, conservation priority aims at identifying species that are facing a high risk of extinction due to their limited distributions, declining population sizes, or sensitivity to environmental changes (Boonman et al. 2024). Notably, priority areas can be based on the relatively precise spatiotemporal dynamics of each species using species distribution models.

The present study aimed to fill the gaps of how future climate and land cover changes may affect distributions, species richness, and turnover rates of narrow‐ and wide‐ranging species across different plant lifeforms in northern China. We hypothesized that: (1) global climate change is likely driving spatiotemporal patterns of plant distributions and altering plant diversity in northern China, with narrow‐ranging species experiencing greater changes than wide‐ranging species; and (2) species composition and richness are more sensitive to climate change in harsh environments than in temperate environments, and, thus, species turnover rates increase linearly with ecological vulnerability. To test these hypotheses, using species distribution models, we analyzed the spatiotemporal distribution patterns of 5111 plant species, encompassing 149 families and 1027 genera from 111,071 occurrence records in northern China. Specifically, we (i) predicted and compared the potential distribution range of narrow‐ and wide‐ranging species across different plant lifeforms under future climate change scenarios; (ii) estimated patterns of species richness and turnover rates, as well as their trends under current and future climates; (iii) determined the main drivers for changes in the potential range and species richness of plants; and (iv) quantified the relationship of species turnover and ecological vulnerability at the grid level to gain insights into biodiversity dynamics and offer guidance for conservation strategies.

2. Material and Methods

2.1. Study Area

Data were collected in northern China (24°–55° N, 68°–127° E) (Figure S1), a region that is highly heterogeneous geographically and ecologically, and includes various topographic landscapes, including mountain ranges, plateaus, basins, and deserts. The topography spans a broad elevation range from −159 to 8797 m, descending gradually from west to east. It also covers a wide range of regions with different levels of aridity (Zomer et al. 2022), including hyper arid (aridity index < 0.03, hereafter referred to as AI), arid (0.03 < AI < 0.2), semi‐arid (0.2 < AI < 0.5), arid semi‐humid (0.5 < AI < 0.65) and humid (AI > 0.65). Annual precipitation ranges from 9 to 3229 mm, annual temperature ranges from −23.2°C to 24.4°C, and the main vegetation types are desert, grassland, and meadow. These diverse characteristics make the region an ideal place to study the response of plant diversity to future climate change.

2.2. Data

2.2.1. Species Distribution Data

The species distribution data were obtained primarily from an online database and field surveys. The Species 2000 China Node (http://www.sp2000.org.cn/) provided the species catalogue of angiosperms and gymnosperms in the study area, encompassing approximately 13,315 species, 1775 genera, and 194 families. Based on the species names in the catalogue, we obtained the distribution information of each species, namely the latitude and longitude (as of May 10, 2022) of the species, using the R packages “rgbif” (Chamberlain et al. 2021) and “BIEN” (Maitner et al. 2018). These two R packages facilitated our access to the Global Biodiversity Information Facility (GBIF; https://www.gbif.org/) and Botanical Information and Ecology Network (BIEN; http://bien.nceas.ucsb.edu/bien/) databases, respectively. As the existing data were concentrated primarily on grassland species, with little information on desert species, we surveyed desert areas specifically (methods described in Yao et al. 2021) to supplement the database.

We excluded duplicates and incorrect records, and species not within the study area or with fewer than five occurrences. Ultimately, we retained 111,071 occurrence records, containing 149 families, 1027 genera, and 5111 plant species. We classified all species into three different lifeforms: annual herbs (n = 485), perennial herbs (n = 3140), and woody plants (n = 1486). Species were further classified into narrow‐ and wide‐ranging based on their niche breadths (Figure S2). Specifically, we first compiled global occurrence records for each species and, based on their geographic coordinates, extracted the annual mean temperature (MAT) and annual precipitation (MAP) from the WorldClim 2.1 database (http://www.worldclim.org/). For each species, temperature and precipitation niche breadths were calculated as the difference between the maximum and minimum values of MAT and MAP across all occurrence localities. To integrate these two dimensions of climatic tolerance, we standardized each variable using Z‐score transformation to remove unit differences and multiplied the standardized values to generate a composite index representing the overall climatic niche breadth. Species were then ranked by their composite index values. The median was selected as a threshold for classification because it provides a simple and widely used criterion in ecological studies (e.g., Newbold et al. 2014). Species below the median were classified as narrow‐ranging (n = 2556; annual herbs = 177, perennial herbs = 1648, woody plants = 731), while those above the median were classified as wide‐ranging (n = 2555; annual herbs = 308, perennial herbs = 1492, woody plants = 755). We chose this approach because it captures species' climatic tolerance in terms of temperature and precipitation and is less influenced by sampling intensity or spatial clustering than classifications based on the number of grid cells occupied by species. To assess the robustness of our classification, we also tested alternative thresholds based on tertiles and quartiles, as well as a method using the number of occupied spatial grid cells. Among these species, 1480 were identified as endemic to China, including 1237 narrow‐ranging and 243 wide‐ranging species. Additionally, 158 species were classified as threatened in the China Species Red List, including 91 narrow‐ranging and 67 wide‐ranging species.

2.2.2. Environmental Data

Climate data were obtained from the WorldClim 2.1 database. We downloaded 19 bioclimatic variables from 1970 to 2000 as current climate data. For future climate data, we selected three different shared socioeconomic pathway (SSP) scenarios, SSP1‐2.6 (sustainable pathways), SSP2‐4.5 (intermediate pathways), and SSP5‐8.5 (traditional fossil fuel‐dominated pathways). SSP1‐2.6 predicts an average global temperature increase of 2.0°C above pre‐industrial levels by 2100; SSP2‐4.5 predicts an average increase of 3.0°C; and SSP5‐8.5 predicts an average increase of 5.0°C. To reduce uncertainties in future climate projections, we acquired 19 bioclimatic variables from seven climate models (CanESM5, CNRM‐CM6‐1, CNRM‐ESM2‐1, IPSL‐CM6A‐LR, MIROC6, MIROC‐ES2L, and MRI‐ESM2‐0) at a spatial resolution of 10′ × 10′ and calculated their averages to predict future climate data.

The land cover data were derived from the Land‐Use Harmonization (LUH2; http://luh.umd.edu) dataset, which provided 12 land cover types globally from the year 850 to 2100 at a spatial resolution of 0.25° × 0.25° (Hurtt et al. 2020). For current land use data, we selected the LUH2 version 2h (850‐2015), and to match the climate data, averages were calculated for 31 years, 1970–2000; for future land use data, we used the LUH2 version 2f under the RCP2.6 SSP1, RCP4.5 SSP2, and RCP8.5 SSP5 scenarios (2015‐2100), and calculated the average for the 20 years from 2081 to 2100. We then merged the 12 land cover types into 5 major categories: forested, non‐forested, crop, urban, and grazing land.

Temperature, precipitation, and elevation ranges were calculated for each 0.45° × 0.45° grid cell using the ArcGIS‐based “Zonal Statistics as Table” tool to quantify environmental heterogeneity. Specifically, the range was defined as the difference between the maximum and minimum values of all pixels within each grid cell. The climate variables (temperature and precipitation) used in this calculation had an original spatial resolution of 10′ × 10′, while elevation data were obtained from DEM90—EarthEnv (http://www.earth env.org/DEM), with a resolution of 90 m. Due to the lack of future elevation data, it was assumed to remain constant under future climate scenarios, following previous studies (Peng et al. 2021).

Given the various resolutions of environmental variable layers, we used the “Resample” tool in ArcGIS to unify the resolution of all layers to 0.45° × 0.45° (approximately 50 × 50 km) employing the BILINEAR resampling method. To mitigate model overfitting due to collinearity among environmental factors, we used a Pearson correlation analysis on 27 factors. This set encompassed MAT_range (annual mean temperature range), MAP_range (annual precipitation range), ELE_range (elevation range), 19 bioclimatic variables, and 5 land cover types (forested, non‐forested, crop, urban, and grazing land). Factors with high correlation coefficients (|r| > 0.8) were excluded (Delgado‐Baquerizo et al. 2017; Katz 2011), and finally, 14 environmental factors were retained for the species distribution model (Figure S3). These included seven climatic variables (MAT, MDR, mean diurnal range, TSN, MTCQ, MAP, PSN: precipitation seasonality, and PDQ: precipitation in the driest quarter), two environmental heterogeneity variables (MAP_range and ELE_range), and five land cover variables (forested, non‐forested, crop, urban, and grazing land).

2.3. Species Distributions Model

We used the “biomod2” R package to model and predict the potential ranges of species for both current and future scenarios. We selected four models from the “biomod2” package, including the generalized linear model (GLM), generalized boosting model (GBM), random forest model (RF), and maximum entropy model (MaxEnt), as they are robust and accurate, and used frequently in ecological studies (Song et al. 2021). To calibrate and evaluate the models, the distribution data for each species was divided randomly into two parts: 80% served as the training dataset, and 20% as the test dataset for validating the model. The process was repeated five times to obtain robust model predictions, and the accuracy of the model was assessed using the True Skill Statistic (TSS) (Allouche et al. 2006). To enhance the precision of the species distribution model and minimize model uncertainty, we selected the model with a TSS > 0.5 for constructing ensemble models for the current period and under the three different future scenarios. Specifically, we used the mean of all selected model predictions for the final ensemble model. We then transformed prediction values (suitability of species) into binary maps using the TSS maximum method, where 0 indicated absence and 1 indicated presence. Species distribution models often overestimate the current and future potential ranges of species since dispersal limitations are not considered. Following previous studies (Kremen et al. 2008; Peng et al. 2021; Song et al. 2021), we used a 200 km buffered minimum convex polygon surrounding the current distribution range of a species to clip the species distribution range predictions. This analysis used the “gConvexHull” and “gBuffer” functions in the R package “rgeos.”

2.4. Data Analyses

2.4.1. Potential Distribution Ranges and Species Richness Patterns

To compare the changes in species distribution range from the present to the future, we used the “BIOMOD_RangeSize” function in the “biomod2” R package. This function enabled us to quantify two key metrics for each species: the percentage change in suitable habitat (CSH) and the percentage loss of suitable habitat (LSH), which were calculated as follows:

$$\mathrm{CHS}=\left(\left[{\text{Area}}_{\text{Future}}-{\text{Area}}_{\text{Current}}\right]/{\text{Area}}_{\text{Current}}\right)\times 100$$ (1)

$$\mathrm{LSH}=100-\left[\left({\text{AreaOverlap}}_{\left(\text{Future},\text{Current}\right)}/{\text{Area}}_{\text{Current}}\right)\times 100\right]$$ (2)

where Area_Future and Area_Current are the areas of future and current habitats. Species with LSH ≥ 30% were identified as threatened species, facing a high risk of extinction (IUCN 2012). The CSH estimates the extent of expansion or contraction of the species distribution range, with positive values representing an increase in suitable habitat areas and negative values indicating a decrease. LSH reflects the extent to which a species may lose its original suitable habitat under future climate change. The larger the LSH, the greater the percentage of its current suitable habitat that will be lost in the future.

By overlaying binary maps of each species' potential distribution range, we generated species richness patterns. The difference in richness between the current and future projections equaled the species richness change, where positive differences denoted increases in richness and negative differences denoted decreases.

2.4.2. Distance and Direction of Shifts in Species Distribution Range

To determine the potential direction and distance of the species shift in distribution range under future climate change scenarios, we determined the changes in the geographical centers (centroids) of each species' distribution between current and future climate scenarios. Specifically, we first used the “gCentroid” function in the R package “rgeos” to determine the centroid of each species distribution in different time periods. Then, we calculated the movement distance and direction of the species using the “distm” and “bearing” functions, respectively, in the R package “geosphere.” Additionally, based on the centroids of all species' distribution at different periods, we used the “Directional Distribution” tool in ArcGIS to draw standard deviation ellipses to further determine the direction of the shift in the species distribution range under future climate and land cover scenarios.

2.4.3. Key Environmental Variables

To detect the key environmental drivers of species distribution, we assessed the importance of each variable for each species using “get_variables_importance” function in the R package “biomod2.” To examine how changes in environmental variables affect LSH, we employed beta regression analysis, which is particularly well‐suited for proportion data constrained within a 0–1 range (Ferrari and Cribari‐Neto 2004). The dependent variable is LSH, while the independent variables are the changes in environmental variables from the present to the future. For each species, the change in environmental factors is represented by the average environmental change in the grids where the species is predicted to be lost in the future. To compare the relative influence of each variable, we standardized each variable using the Z‐score transformation.

2.4.4. Identification of Priority Conservation Areas and Species

We first calculated the species lost, gained, and turnover rate in each grid cell, as follows (Thuiller et al. 2005):

$$\text{Species loss rate}=L/\mathrm{SR}\times 100$$ (3)

$$\text{Species gained rate}=G/\mathrm{SR}\times 100$$ (4)

$$\text{Species turnover rate}=\left(L+G\right)/\left(\mathrm{SR}+G\right)\times 100$$ (5)

where L is the number of species lost within a grid cell (equivalent to the number of species migrating out), G is the number of species gained within a grid cell (equivalent to the number of species migrating in), and SR is the current species richness within a grid cell.

Using the “Hot Spot Analysis (Getis‐Ord Gi*)” tool in ArcGIS, we identified areas with particularly fast and slow species loss rates, gain rates, and turnover rates. Subsequently, the species turnover rate layer was overlaid with the protected area layer to pinpoint priority areas for conservation or regions with conservation gaps. Furthermore, we explored the relationship between species turnover rates and the ecological vulnerability index, which is used to quantify the susceptibility of ecosystems to environmental change or disturbances. A detailed assessment of the ecological vulnerability index is provided in Appendix S1. To identify priority species, we compiled a list of species with LSH greater than 80% under the three climate scenarios. Additionally, we focused on the CSH and LSH of species listed as threatened in the China Species Red List.

3. Results

3.1. Model Performance

The TSS values ranged from 0.65 to 1, with most above 0.9 (Figure S4). This range indicated the credibility of the ensemble model in predicting species distribution in response to climate and land cover changes.

3.2. Projected Trends in Species Distribution Ranges Under Climate Scenarios

Using the SSP1‐2.6, SSP2‐4.5, and SSP5‐8.5 scenarios, 64%–70% of the species expanded their distribution ranges (CSH > 0), while 28%–33% contracted their ranges (CSH < 0) (Figure 1a). Among species with expanding ranges, wide‐ranging species slightly outnumbered narrow‐ranging species, particularly annual herbs and woody plants (Figure 1b,d). In contrast, among contracting ranges, narrow‐ranging species outnumbered wide‐ranging species, especially when CSH ≤ 50% (Figure 1). Overall, the range contraction of narrow‐ranging species was greater than that of wide‐ranging species, while their range expansion was less (Figure S5). For LSH, 61%–69% of the species had an LSH below 30%, while 31%–39% faced a high risk of habitat loss (LSH ≥ 30%) across the SSP1‐2.6, SSP2‐4.5, and SSP5‐8.5 scenarios (Figure 2). Narrow‐ranging species consistently displayed a greater LSH than wide‐ranging species (Figure S6). This was particularly evident when LSH exceeded 50%, with a larger percentage of narrow‐ranging species that were likely to reduce their original habitat area (Figure 2). Our results were also robust to the use of alternative thresholds based on tertiles and quartiles, as well as to a method based on the number of occupied spatial grid cells (Figures S7 and S8). The classifications based on tertiles and quartiles yielded results highly consistent with those based on the median (Figure S7). Similarly, when species were classified using the number of occupied grid cells, the differences in CSH and LSH between narrow‐ and wide‐ranging species were consistent with those based on climatic niche breadth (Figure S8).

FIGURE 1.

The percentage of plant species changing habitat suitability (CSH) in the future under three climate and land cover change scenarios (SSP1‐2.6, SSP2‐4.5, SSP5‐8.5). (a) All plants. (b) Annual herbs. (c) Perennial herbs. (d) Woody plants. Border color represents narrow‐ranging species, and fill color represents wide‐ranging species.

FIGURE 2.

The percentage of plant species losing suitable habitat (LSH) in the future under three climate and land cover change scenarios (SSP1‐2.6, SSP2‐4.5, SSP5‐8.5). (a) All plants. (b) Annual herbs. (c) Perennial herbs. (d) Woody plants. Border color represents narrow‐ranging species, and fill color represents wide‐ranging species.

The relationship between CSH and LSH revealed that LSH generally increased as the suitable habitat contracted (Figure 3). Notably, even when CSH was positive, many species still experienced increases in LSH, particularly when CSH exceeded 30%. This suggests that species may lose portions of their original habitat despite an overall range expansion. This pattern was especially pronounced for narrow‐ranging species under the SSP5‐8.5 scenario (Figure 3a).

FIGURE 3.

The relationship between the percentage of potential distribution range changes (CSH) and the percentage of suitable habitat loss (LSH) in the future under three climate and land cover change scenarios (SSP1‐2.6, SSP2‐4.5, SSP5‐8.5). (a) Narrow‐ranging. (b) Wide‐ranging.

Species migration was directed primarily towards the west, northwest, north, and northeast, with relatively few species migrating southwest (Figure 4a,b). Migration distances were greater under SSP5‐8.5 than SSP1‐2.6, but narrow‐ and wide‐ranging species did not differ across plant lifeforms (Figure S9). Overall, as climate scenarios intensified, species distributions shifted progressively northwestward, indicating a trend toward higher latitudes and elevations under future climate change (Figure 4c).

FIGURE 4.

The migration distances and directions of species. (a) Migration distances and directions of narrow‐ranging species. Colors indicate migration distance; numbers denote the number of species. (b) Migration distances and directions of wide‐ranging species. (c) Main migration directions of all species. The background map shows elevations. Ellipses are standard deviation ellipses generated from the centroids of species' potential distributions under different climate scenarios using the “Directional Distribution” tool in ArcGIS. They summarize spatial characteristics such as central tendency, dispersion, and directional trends. The blue arrow indicates the dominant migration direction. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

3.3. Variations in Species Richness

Species richness in both narrow‐ and wide‐ranging species, across different lifeforms, displayed a consistent pattern under different climate scenarios. Species richness was greatest in the southeastern part of the study area (Figure S10). Narrow‐ranging species were particularly dominant in cold and arid regions, in particular in the northern Tibetan Plateau and the Tengger Desert (Figure 5). However, narrow‐ranging species, particularly those in the western edge of the Tibetan Plateau and the northeastern part of the study area, are predicted to experience a substantial decline in species richness in the future (Figure 6a,c,e,g). In contrast, an overall increase in species richness is projected across most of the study areas, with wide‐ranging species expected to experience the most notable gains (Figure 6b,d,f,h). Of the three scenarios, the SSP5‐8.5 scenario is predicted to have the strongest positive impact on species richness.

FIGURE 5.

Spatial distribution patterns of the percentage of species richness for narrow‐ and wide‐ranging species in the current period and under three future climate and land cover change scenarios (SSP1‐2.6, SSP2‐4.5, SSP5‐8.5). (a) Narrow‐ranging species. (b) Wide‐ranging species. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

FIGURE 6.

Future trends in plant species richness patterns of narrow‐ and wide‐ranging species across different lifeforms under three climate and land cover change scenarios (SSP1‐2.6, SSP2‐4.5, SSP5‐8.5). Yellow indicates a decrease and green indicates an increase. The deeper the color, the greater the magnitude of change. The numbers in the figure indicate the percentage of the study area covered by regions with an increase (green) or decrease (orange) in richness. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

3.4. Effects of Climate and Land Cover Changes on Species Distribution

Overall, climatic factors will have a greater impact on plant species distribution than land cover factors. For both narrow‐ and wide‐ranging species, the primary climatic drivers of species richness will remain similar (Figure 7a,c). Among plant lifeforms, annual and perennial herbs will be influenced mainly by TSN, MAP, and PSN, while woody plants will be more affected by TSN, MAP, and MTCQ (Figure S11). Specifically, the changes in MAT and MTCQ will have a significant positive impact on LSH. The changes in MAP will affect LSH positively for narrow‐ranging species, but negatively for wide‐ranging species. In terms of climate variability, MDR and TSN are correlated positively, but PSN is correlated negatively with LSH. Changes in forested, crop, and urban land covers may lead to habitat fragmentation, thereby exacerbating the risk of species loss (Figure 7b,d).

FIGURE 7.

Importance of environmental factors and relationships between the percentage of suitable habitat loss (LSH) of the species and changes in the environmental variables based on beta regression models. (a, b) Narrow‐ranging species. (c, d) Wide‐ranging species. Δ: the change in environmental variables between the current and future scenarios; Crop: crop land; ELE_range: elevation range; Forested: forest land; Grazing: grazing land; MAP: annual precipitation; MAP_range: annual precipitation range; MAT: annual mean temperature; MDR: mean diurnal range; MTCQ: mean temperature of coldest quarter; Non‐forested: non‐forest land; PDQ: precipitation of driest quarter; PSN: precipitation seasonality; TSN: temperature seasonality; Urban: urban land.

These differences in the effects of climate factors also highlight the importance of species' environmental tolerance in shaping their distributions. Species with expanding distribution ranges tend to adapt to a wider range of temperature and precipitation changes (Figure S12). In contrast, species with contracted ranges, although adaptable to a broader range of temperatures, exhibit weak tolerance to precipitation changes, which limits their distribution under future climate scenarios. In addition, species with a narrow climatic niche breadth tend to have a large LSH, meaning that they are more likely to lose a large portion of their suitable habitats (Figure S13).

3.5. Identification of Priority Conservation Areas and Protected Species

The southern part of the study region, characterized by rich biodiversity, requires enhanced protection; whereas areas with fast species turnover rates, including the Taklimakan Desert, Tengger Desert, Qilian Mountains, and the central and western Tibetan Plateau, were identified as critical regions for conservation (Figure 8a–c). In the Taklimakan Desert and central Tibetan Plateau, the fast species turnover rate was caused mainly by the fast rate of gained species (Figure S14d–f), whereas, in the Qilian Mountains and western Tibetan Plateau, turnover was attributed mainly to the fast rate of lost species (Figure S14a–c). Species turnover rates exhibited a significant positive relationship with ecological vulnerability under all scenarios (Figure 8d–f). These fast‐turnover regions were often located in mountainous and desert ecosystems; however, China's existing protected areas cover only a fraction of these ecologically important regions, and the areas not covered by these protected areas are mainly located in the Taklamakan Desert, Badain Jaran Desert, Qaidam Basin, and the eastern edge of the Tibetan Plateau. Priority conservation areas located outside the boundaries of protected areas exhibited greater (p < 0.001) ecological vulnerability than other regions (Figure 8g,h).

FIGURE 8.

Species turnover rate in the study area in the future under three climate and land cover change scenarios (SSP1‐2.6, SSP2‐4.5, SSP5‐8.5). (a–c) Species turnover rate at the grid level. Red dots represent areas with significantly fast species turnover rates; blue dots represent areas with significantly slow species turnover rates; those that are not significant are not displayed. (d–f) Relationship between species turnover and the ecological vulnerability index. (g) Priority conservation areas but not covered by nature protected areas. Gray represents mountains, green represents protected areas, and yellow represents deserts or sand. (h) The ecological vulnerability of priority conservation areas located outside the boundaries of protected areas (unprotected areas) and other areas. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

Approximately 30%–40% of the species are projected to lose more than 30% of their suitable habitats across all scenarios. A total of 187 species, including 73 endemic species such as Bupleurum commelynoideum var. flaviflorum, Carex hohxilensis, Gentiana spathulifolia, and Rhododendron capitatum, are projected to experience an LSH exceeding 80% across all three climate scenarios (Table S1). Additionally, 158 threatened species listed in the China Species Red List were identified, with 66%–78% expected to expand (CSH > 0) and 22%–34% to contract (CSH < 0) their distributions (Table S2). Furthermore, 28 of 97 species occurring in high‐elevation and northwestern marginal areas are projected to lose over 80% of their suitable habitats, indicating increased extinction risk in geographically or topographically constrained regions. Ten representative species are presented in Figure S15, and the full list of these 97 species is provided in Table S3.

4. Discussion

The distribution range and species richness in northern China are projected to increase in the future. This trend is influenced primarily by TSN, MAP, and MTCQ, which indicate that temperature and precipitation play important roles in shaping the species distribution and richness patterns in northern China. In comparison to climatic variables, land cover factors do not have a large impact on species distribution and richness, perhaps because land cover is not expected to change markedly in the study area in the future (Figure S16). Among climate factors, TSN can enhance the phenotypic plasticity of plants, thereby improving adaptability under climate change, and expanding their distribution range (Molina‐Montenegro and Naya 2012). However, TSN also affects physiological processes such as plant growth, reproduction, and dormancy. In environments with large changes in TSN, plants need to exhibit strong adaptability and resilience to cope with extreme temperature fluctuations (Hawkins et al. 2014; Huang et al. 2021). The MTCQ is related to the plants' cold tolerance. Many plants have specific temperature thresholds for cold tolerance, and temperatures below these thresholds can cause cellular freezing or physiological dysfunction, thus affecting plant survival. For plants with poor cold tolerance, their distribution range is often constrained by the minimum temperature (Šímová et al. 2011; Wu et al. 2018). MAP directly affects the water source needed for plant growth. Regions with insufficient precipitation may face water scarcity, limiting plant distribution (Korell et al. 2021).

4.1. Change in Species Distribution and Potential Causes

In northern China, it was predicted that about 70% of plant species would expand their distribution ranges. However, this finding is not in agreement with Dong, Gong, et al. (2024), who reported that 42%–55% of vascular plants in Inner Mongolia are predicted to contract their suitable distribution areas. Dong, Gong, et al. (2024) focused on rare and endangered vascular plants, which typically have narrow geographic distributions and are particularly sensitive to climate change. In contrast, the present study included many wide‐ranging species that are tolerant of climate change. Additionally, the study area is larger, and future projections suggest increases in MAT, MTCQ, MAP, and PDQ are expected (Figure S17). These factors may contribute to the distribution range expansion. For instance, the rise in temperature would enable the survival of many species that would not be able to do so under the former cold conditions. Consequently, the warming climate may enhance the survival of certain species, such as woody plants (Guo et al. 2023), leading to an expansion of their distribution range. Our results further displayed a hump‐shaped relationship between species distribution ranges and MAT, with both low and high temperatures limiting distribution. The distribution range reached its maximum at the optimal temperature (Figure S18a). Furthermore, increased precipitation not only improves the soil water availability in arid or barely suitable areas, transforming them into more favorable habitats, but also offsets the physiological limitations caused by increased temperature (Hu et al. 2022). Species distribution also increased with MAP in regions with low precipitation (Figure S18b). As a result, species previously constrained by insufficient water availability can now thrive in these newly suitable regions. For species under survival pressure in arid conditions, increased precipitation can alleviate water‐related stress, enhancing survival and reproductive success, and promoting species expansion (Reich et al. 2022).

However, not all species benefit from climate change. Range contraction for species can be attributed to rising temperatures that exceed the optimal tolerance level (Dolezal et al. 2021). Additionally, when comparing the environmental characteristics of species reducing their ranges to those expanding their ranges, the former species generally inhabited higher elevations (Figure S19). This difference between species supported the “mountain‐top extinction hypothesis,” which posited that as the climate warms, species on mountaintops are at a risk of extinction due to their limited adaptation to warmer temperatures (Dullinger et al. 2012; Liang et al. 2018; Baumbach et al. 2021). These species, adapted to the cooler climates of high elevations, move to even higher elevations. However, being already at the summit, they cannot move to higher ground, resulting in a gradual loss of viable habitat (Dirnböck et al. 2011).

Under all three scenarios (SSP1‐2.6, SSP2‐4.5, SSP5‐8.5), narrow‐ranging species exhibited greater changes in CSH and LSH than wide‐ranging species. These results clearly indicated that narrow‐ranging species will face greater survival challenges under future climate change than wide‐ranging species (Slatyer et al. 2013), which supports our first hypothesis. This is likely because narrow‐ranging species generally display a more limited niche breadth and lower adaptability to environmental changes, leading to greater habitat alterations and losses. In contrast, wide‐ranging species, with their broader niche breadth, can adapt to a wider range of environmental conditions, enabling them to survive and thrive across a broader spectrum of climatic scenarios (Xu et al. 2019; Tang et al. 2021).

In addition to changes in distribution ranges, the direction of species migration was evident. Most species will migrate primarily northwest, although some will also expand their distribution ranges towards the southwest, west, north, and northeast. We speculate that the range shifts can be attributed to the following factors: Firstly, climate warming drives many species northwestward. As temperatures rise, species tend to move toward cooler regions, which are found near the poles or at higher elevations (Chen et al. 2011). The northwestward shift aligns with the broader pattern of species moving toward cooler, more suitable habitats as the climate warms. Moreover, this direction coincides with the region's natural topography, where mountainous areas and higher altitudes are more prevalent in the northwest, offering favorable conditions for species sensitive to a rise in temperature (Yu et al. 2019). Secondly, local geographical and ecological factors contribute to species shifting in other directions, such as southwest, west, north, and northeast. These shifts could be driven by changes in regional temperature and precipitation patterns or the emergence of specific microhabitats that become more favorable under new climate conditions (Lenoir and Svenning 2015; Liang et al. 2018).

4.2. Species Richness Under Climate and Land Cover Changes

The present study indicated that future climate change and land cover modifications will alter the species richness pattern in northern China only slightly. Specifically, the projected future pattern will closely resemble the current one, with greater species richness in the southeast and less in the northwest. Notably, narrow‐ranging species will be more prevalent in cold and arid regions, such as the northern Tibetan Plateau and the Tengger Desert. This prevalence is due primarily to their evolutionary specialization, allowing them to adapt to extreme conditions, while the low competition pressure and ecological isolation in these environments will also promote their survival and reproduction (Wiens and Graham 2005). However, our projections indicated a decrease in the proportion of narrow‐ranging species richness in these regions in the future. This result is consistent with the findings of Xu et al. (2019, 2023), who stated that widespread species tend to be “winners” during biodiversity changes, while narrowly distributed species are often “losers.” The reason could be that the TSN in this region is projected to increase under future climate scenarios (Figure S17). This increase in TSN will result in greater temperature stratification, which will require plant species to tolerate a broader range of temperature variations for dispersal, posing a challenge for narrowly distributed species (Qian et al. 2022). Consequently, narrow‐ranging species in arid and cold regions may ultimately be replaced by wide‐ranging species (Newbold et al. 2018).

Although species richness among narrow‐ranging species is expected to decline in the future, the overall species richness across the study area is projected to increase. This is consistent with previous research, which used random forest models to demonstrate that species richness in northern China is expected to increase (Sun et al. 2021). We also found that species richness in the study area generally increases with MAT, but this effect diminishes at higher temperatures (Figure S18c). Additionally, species richness in regions with low precipitation (MAP < 1000 mm) tends to increase with precipitation, suggesting that future increases in MAP could further enhance species richness in northern China (Figure S18d). This aligns with previous studies (Korell et al. 2021; Chen et al. 2023; Kou et al. 2023), which predicted that moderate warming and increased humidity could enhance plant species richness in arid regions. In this context, annual plants typically exhibit high reproductive rates and strong environmental adaptability, enabling them to adjust quickly to their life cycles and accommodate new climatic conditions. Warm and moderately humid environments not only facilitate the germination of annual herb seeds and the survival of seedlings but also extend the growing season, providing more time for growth and reproduction, thereby increasing their richness (Facelli et al. 2005; Chen et al. 2009). For perennial plants, their deep and extensive root systems confer high resource use efficiency, enabling them to be competitive and thrive even in conditions of drought and fluctuations in precipitation (Vico et al. 2016). Woody plants have a strong tolerance to temperature variations, and climate warming may enable some species, previously restricted by low temperatures, to expand into broader areas (Saintilan and Rogers 2015). Additionally, the rise in atmospheric CO₂ concentration accompanying climate warming provides more resources for photosynthesis, further promoting the growth and increase in richness of woody plants (Ainsworth and Long 2005).

4.3. Implications for Plant Conservation Under Climate Change

Identifying and prioritizing conservation efforts in regions with fast species turnover rates is crucial under climate change scenarios. Areas such as the Tibetan Plateau, Qilian Mountains, Taklimakan Desert, and Tengger Desert not only exhibit fast species turnover rates but also demonstrate high ecological vulnerability, further validating our second hypothesis. The ecological vulnerability of these regions makes them more sensitive to climate change. Thus, conservation strategies should focus on preserving these mountain and desert ecosystems and maintaining biodiversity by establishing or expanding protected areas to mitigate the effects of climate change.

Species with greater than 30% LSH face survival threats, as habitat reduction can impact their population (Chowdhury 2023). Among these, 193 species with LSH greater than 80% across all climate scenarios, including 73 that are endemic to China, highlight the urgency of protection efforts. These species are particularly vulnerable due to their limited ability to adapt to rapid environmental change. In particular, species distributed in high‐elevation and northwestern marginal areas also face heightened extinction risk due to geographic or topographical constraints. Furthermore, some threatened species listed in the Red List with LSH less than 30% may have relatively stable habitats; however, risks remain due to environmental or anthropogenic factors, particularly for those with specialized habitat requirements. Narrow‐ranging species in the Red List are especially sensitive to environmental changes, making them more susceptible to future habitat shifts (Liu et al. 2023).

Although some species may expand their distribution ranges under future climate scenarios, this should not remove conservation concerns. Despite the potential expansion of their distribution ranges, these species are likely to increase LSH, resulting in the loss of some suitable habitats, which raises concerns about whether these species can disperse or migrate successfully to new habitats. Therefore, species with potentially expanding distribution ranges should not be considered completely safe from threats, as habitat expansion does not guarantee survival since factors such as competition and dispersal limitations may affect their persistence (Engler et al. 2009). For species with high LSH, in situ conservation should be prioritized as an urgent strategy to ensure their survival with climate change.

The future expansion of woody plants will be due primarily to the expansion of shrubs (Figures S20 and S21). The topic of shrub expansion in arid and semi‐arid areas remains controversial. On the one hand, it is viewed as a signal of grassland degradation. Previous studies indicated that shrub expansion can diminish the richness, abundance, and biomass of herbaceous plants in arid regions, with drought exacerbating these negative impacts. On the other hand, shrub expansion increases the carbon and nitrogen content in soils and enhances the water and nutrient storage capacities of the ecosystem (Zhou et al. 2019). In colder regions, shrub encroachment diminished herbaceous biomass and pathogen load; however, in warmer regions, shrubs did not affect the pathogen load (Dang et al. 2025). Regular biodiversity surveys are essential to track shifts in plant communities in northern China under future climate change. The establishment of a comprehensive database would enhance our ability to predict how these changes may impact ecosystem functions.

4.4. Limitations of the Study

To enhance the species distribution models, we employed an ensemble model approach. However, our models still had limitations. Firstly, the models assumed niche conservatism, meaning that species' niches remain unchanged and do not evolve over time in response to climate change. Secondly, interactions between species and other potential influencing factors were not fully considered in the models. Previous studies attempted to integrate species traits (Benito Garzón et al. 2019) or phenological characters (Peng et al. 2024) into species distribution models, achieving some progress in improving model prediction accuracy. Further studies are warranted in incorporating these factors to better describe the complexity of ecosystems. For example, species traits could be examined more closely to reveal potential adaptive mechanisms in response to environmental changes. These efforts would not only enhance the reliability of species distribution models but also offer more accurate scientific guidance for biodiversity conservation and ecosystem management.

5. Conclusions

This study predicted how climate change impacts the spatiotemporal patterns of species distribution, richness, and turnover rates in northern China, emphasizing the relationship between ecological vulnerability and species turnover. It was predicted that the distribution ranges of most species will expand, and species richness will increase. However, the fast species turnover rate in mountains and deserts highlights the need for conservation efforts. These regions, with their fast species turnover rate and ecological vulnerability, are expected to experience more dramatic shifts in species composition. Consequently, special attention should be provided for species with high LSH or narrow ranges. The present study provides important insights into the spatiotemporal patterns of plant species distribution and richness and underscores the urgency of incorporating climate change projections into biodiversity conservation strategies to safeguard vulnerable ecosystems.

Author Contributions

Ying Sun: data curation, formal analysis, investigation, methodology, validation, visualization, writing – original draft, writing – review and editing. Yan Deng: data curation, formal analysis, methodology, validation, visualization, writing – original draft, writing – review and editing. Shuran Yao: data curation, investigation, writing – original draft. Yuan Sun: data curation, investigation, writing – original draft. Abraham Allan Degen: methodology, validation, writing – review and editing. Longwei Dong: methodology, validation, writing – original draft. Jiali Luo: data curation, investigation, writing – original draft. Shubin Xie: methodology, validation. Qingqing Hou: methodology, validation. Dong Tang: methodology, validation. Yuzhen Sun: methodology, visualization. Junlan Xiong: data curation, methodology. Jie Peng: validation, writing – review and editing. Weigang Hu: investigation, writing – review and editing. Jinzhi Ran: conceptualization, data curation, investigation, methodology, supervision, validation, writing – original draft, writing – review and editing. Jianming Deng: conceptualization, data curation, funding acquisition, investigation, methodology, resources, supervision, visualization, writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Sun, Y. , Deng Y., Yao S., et al. 2025. “Distribution Range and Richness of Plant Species Are Predicted to Increase by 2100 due to a Warmer and Wetter Climate in Northern China.” Global Change Biology 31, no. 7: e70334. 10.1111/gcb.70334.

Data Availability Statement

The data that support the findings of this study are openly available in Dryad at https://doi.org/10.5061/dryad.zpc866tmp. Species occurrence data were obtained from the Global Biodiversity Information Facility (GBIF) at https://doi.org/10.15468/dl.4fzesm and the Botanical Information and Ecology Network (BIEN) at https://bien.nceas.ucsb.edu/bien/ (version 4.2). Climate data were obtained from WorldClim 2.1 at https://www.worldclim.org/. Land cover data were obtained from the Earth System Grid Federation at https://doi.org/10.22033/ESGF/input4MIPs.1127, https://doi.org/10.22033/ESGF/input4MIPs.1661, https://doi.org/10.22033/ESGF/input4MIPs.1886, and https://doi.org/10.22033/ESGF/input4MIPs.1662.