The study area, located between 31°42′-53°23′N and 73°40′-126°04′E, covers six provinces and autonomous regions in Northwest China, i.e., Xinjiang Uygur Autonomous Region (Xinjiang), Gansu Province (Gansu), Qinghai Province (Qinghai), Shaanxi Province (Shaanxi), Ningxia Hui Autonomous Region (Ningxia), and Inner Mongolia Autonomous Region (Inner Mongolia) (Fig. 1). Topography of this region is predominated by plateaus, mountains, and basins. This region has the widest expanse of saline-alkaline land in China, notably in Xinjiang, Qaidam Basin of Qinghai, Hexi Corridor of Gansu, and western Inner Mongolia (Wang et al., 2021). Situated in the northwestern inland area of China, the region is characterized by the arid and semi-arid climate. The average annual temperature ranges from 0.00°C to 16.00°C, with a trend of increasing from northeast to southwest. The annual precipitation is under 200.00 mm in arid areas; whereas in semi-arid areas, it ranges from 200.00 to 400.00 mm, decreasing gradually from east to west. Due to the significant disparity between evaporation and precipitation throughout the year and unreasonable agriculture irrigation methods utilized in the study area, salt accumulation occurs on the ground surface, which results in serious soil salinization.

In addition to field surveys of the selected six halophytic plant species in this study during June-September in 2023, occurrence information was collected from the Chinese Virtual Herbarium (http://www.cvh.org.cn), Specimen Resources Sharing Platform for Education (http://nh.scu.edu.cn/list/locality), Global Biodiversity Information Facility (http://www.gbif.org), and National Specimen Information Infrastructure (http://www.sii.org.cn/). A total of 1432 geographical distribution points were obtained for the six halophytic plant species (427 for S. salsa, 111 for S. europaea, 307 for K. foliatum, 286 for H. caspica, 151 for H. glomeratus, and 150 for H. strobilaceum). We removed the duplicate and unclear records according to the software requirements for building an SDM. To reduce the impact of sampling deviation on model accuracy (Warren et al., 2010), we used the ENMTools software (https://github.com/danlwarren/ENMTools) to trim the distribution points. Finally, 645 geographical distribution points were selected (77 for S. salsa, 69 for S. europaea, 167 for K. foliatum, 167 for H. caspica, 99 for H. glomeratus, and 66 for H. strobilaceum). The species distribution point format was converted into ''.csv'' format to ensure that the model can run smoothly (Li et al., 2020). Importantly, the accuracy of the MaxEnt modeling findings employed in this study was unaffected by fluctuations in occurrence data (Xiong et al., 2019; Fang et al., 2021). ArcGIS 10.8 (ESRI) was used to build the grids (1 km×1 km for each) in the study area, ensuring only one record in each grid (Fig. 1).

In this study, 40 environmental variables were selected to construct the MaxEnt model, including 19 bioclimatic factors, 3 topographic factors, and 18 soil factors (Table S1). The 19 bioclimatic and altitude factors were collected from WorldClim (http ://www.worldclim.org/). The slope and aspect data were calculated using the slope and aspect tools in ArcGIS 10.8 software. The WorldClim dataset provides version 2.1 of bioclimatic data with a spatial resolution of 2.5′, covering the period from 1970 to 2000, as well as future bioclimatic data (note that the bioclimatic data in the 2050s, 2070s, and 2090s represent the average values during the periods of 2041-2060, 2061-2080, and 2081-2100, respectively). We employed predictions from a high-resolution version of the Beijing Climate Center Climate System Model (Zhao et al., 2021), which is part of the Coupled Model Intercomparison Project Phase 6 conducted by the Intergovernmental Panel on Climate Change (IPCC) (Popp et al., 2017). The predictions were based on the scenarios of SSPs reflecting future climate change, indicating that the model is well-suited for studies in China (Yang et al., 2016; Wu et al., 2019). In accordance with the Sixth Assessment Report of the IPCC published in 2021 (Lynn and Peeva, 2021), SSP245 and SSP585 respectively indicate the moderate and highest forcing scenarios (radiative forcing 4.5 and 8.5 W/m² in 2100, respectively) (Thomson et al., 2011; Xian et al., 2022). The soil factor data were obtained from the Harmonized World Soil Database v1.2 (https://www.fao.org/) (FAO, 2012), and A Big Earth Data Platform for Three Poles (https://poles.tpdc.ac.cn/) was used to download Chinese soil organic matter data with a resolution of 1 km (Wei et al., 2013). All the above 40 environmental variables are listed in Table S1. Finally, using ArcGIS 10.8 Conversion tools, all environmental variables were converted to ''ASCII'' format, and their spatial resolutions were changed.

The selection of environmental variables is the key to the operation of the mode, directly influencing both niche analysis results and the prediction of the eventual possible distribution area. The evaluation of the model results can be improved by removing the multiple collinearities of environmental variables. Numerous studies have reported that the multi-ollinearity between the selected environmental variables can cause the model to become overfit and impact the assessment of the model results (Sheppard, 2013; Feng et al., 2019; Wang et al., 2024). Therefore, to ensure the mutual independence of environmental variables, we conducted Pearson's correlation analysis on the 40 environmental variables included in this study. Subsequently, the 40 environmental variables were filtered by the MaxEnt model, eliminating factors with a contribution rate lower than 1.00%. In addition, variables with scores lower than 0.10 in the regularization training gain were removed, as low contributing environmental variables can lead to a decrease in the accuracy of the model's predictions. Finally, if the Pearson's correlation coefficient between the two environmental variables has an absolute value exceeding 0.80 (|r|>0.80), only the variables exhibiting a higher contribution rate were selected following the approach proposed to reduce the occurrence of model overfitting (Cao et al., 2016). Finally, we selected 16 environmental variables with less Pearson's correlation (|r|<0.80) (Fig. 2; Table S2). As the environmental parameters of the final model, they were entered into the MaxEnt software for predicting the ranges and habitats of the six halophytic plant species.

Plant habitat suitability was modeled using a selected MaxEnt model (MaxEnt version 3.4.1; https://biodiversityinformatics.amnh.org/open_source/maxent/) (Elith et al., 2006; Phillips et al., 2009). The model training process used 75.00% of the species distribution records, and the remaining portion was allocated for model testing. The resilience of the model was assessed using 20 replicates. This involved setting 5000 maximum iterations and utilizing 10,000 background samples to achieve the optimal solution. The averages of the results were measured, with other parameters setting at default values. The model's accuracy was evaluated using the area under the receiver operating characteristic curve (AUC) (Na et al., 2018). Previous study has validated that AUC is a reliable measure for both calibrating and validating the model's accuracy (Elith et al., 2011). The AUC value ranges from 0.50 (indicating that the model's accuracy is equivalent to that of a random model) to 1.00 (representing an ideal state where the model's predictions perfectly align with the actual outcomes). An AUC value close to 1.00 indicates the superior prediction ability of the model (Phillips et al., 2006). The widely adopted standard is that the AUC value falls between 0.00 and 1.00, with 0.50-0.60, 0.60-0.70, 0.70-0.80, 0.80-0.90, and >0.90 indicating failed, poor, general, good, and excellent performance, respectively (Srivastava et al., 2019; Tu et al., 2021; Yin et al., 2023). Furthermore, appropriateness maps were created by computing the potential habitat suitability index (varying from 0.00 to 1.00) for the species distribution using the logistic output of the MaxEnt model. On this scale, the index value of 0.00 indicates a completely unsuitable habitat, and the higher the index value, the greater the suitability. In ArcGIS 10.8 software, the suitability index was divided into four categories using natural breaks: 0.00-0.20, 0.20-0.40, 0.40-0.60, and 0.60-1.00 for non-suitable, low suitable, medium suitable, and high suitable areas, respectively (Zhang et al., 2019; Ye et al., 2021). The assessment of the distribution model's environmental predictor contributions was conducted by analyzing percentage variable contributions and employing the Jackknife program (Xiong et al., 2020). This test was employed to ascertain the impact of every independent variable within the model. The suitable range of environmental variables for the potential distribution of six halophytic plant species was identified (Table S3). On this basis, the possible changes and distribution dynamics of a suitable habitat area for plant species in the future were calculated. The predicted results of plant species for different periods were converted into binary distribution files. Further, the change trend of suitable areas was computed through the SDM toolbox, and the shrinking, expanding, and stable regions for plant species distribution were obtained. The Centroid Shift toolbox in the SDM toolbox was used to obtain the centroids of the suitable habitat areas. The migration routes of high suitable habitats can be obtained by observing the movement of centroids in high suitable habitats (Cong et al., 2020) and comparing the future centroid with the existing centroid (Brown, 2014).

The MaxEnt model was employed to forecast the possible spread of six halophytic plant species in various timeframes, both in the current and in the future. The average AUC value was >0.80 (Table 1); therefore, the model's predictions were considered to be of good quality. This study accurately and reliably simulated the suitable ranges of six halophytic plant species under various climate scenarios, employing the MaxEnt model.

The average contribution rates and suitability ranges of environmental variables (Tables S2 and S3) indicated that precipitation seasonality (coefficient of variation) (Bio15), precipitation of the warmest quarter (Bio18), and soil factor exchangeable Na⁺ (Na) exhibited the highest contribution rates (23.15%, 13.22%, and 11.68%, respectively). Among the temperature variables, the average contribution of the mean temperature of the warmest quarter (Bio10) contributed at the highest level, reaching 4.90%. There were significant differences in the contribution of the same environmental variable to each plant species. For example, the contribution rates of Bio15 for H. glomeratus, H. strobilaceum, and H. caspica were 44.00%, 32.00%, and 23.40%, respectively. The suitable ranges of Bio15 for these three species were 23.26-70.13, 23.26-68.34, and 23.22-76.41 mm, respectively. However, the contribution rates of Bio15 for S. salsa and K. foliatum were only 1.70% and 2.00%, respectively. The suitable ranges of Bio15 for S. salsa and K. foliatum were 82.27-111.27 and 37.17-80.05 mm, respectively. Additionally, Bio10 was a secondary factor affecting the distribution of K. foliatum, with a suitable range from -10.33°C to -1.82°C. In addition, among the soil factors, Na was the key factor for S. salsa, S. europaea, H. caspica, and K. foliatum, and the suitable range was from 0.11 to 0.17 cmol/kg.

The Jackknife test results indicated that when the variables were used alone, Bio15 exhibited the greatest gain for H. strobilaceum, H. glomeratus, and H. caspica, and Bio18 exhibited the greatest gain for K. foliatum (Fig. 3). These variables were considered to have the most information about the species distribution. The soil factor Na exhibited a higher gain than other variables for S. salsa and S. europaea.

According to the prediction of the MaxEnt model, different plant species exhibited different characteristics in terms of the area and location of suitable habitats (Fig. 4). The total suitable habitat area (the sum of low, medium, and high suitable areas) for S. salsa was 92.60×10⁴ km²; the medium and high suitable habitats were primarily distributed in central Ningxia, southern Xinjiang, Shaanxi, Inner Mongolia, and parts of Gansu (Fig. 4a). The total suitable habitat areas for S. europaea and H. caspica were 112.00×10⁴ and 118.00×10⁴ km², respectively, predominantly located in Xinjiang, Inner Mongolia, and northern Ningxia (Fig. 4b and e, respectively). The minimum distribution of the suitable habitat area for H. strobilaceum was approximately 47.40×10⁴ km², and the medium and high suitable habitats were predominantly located in Xinjiang and sporadically distributed in Gansu and Inner Mongolia (Fig. 4c). The total suitable habitat area for H. glomeratus was 99.10×10⁴ km²; the medium and high suitable habitats were predominantly situated in Xinjiang and sporadically distributed in Inner Mongolia (Fig. 4d). The maximum total area of suitable habitat for K. foliatum was approximately 150.00×10⁴ km², mainly in Xinjiang, northern Ningxia, central Gansu, central Qinghai, and southwestern Inner Mongolia (Fig. 4f).

Based on the circumstances of SSP245 and SSP585, the suitable habitat areas for different halophytic plant species in the future (2050s, 2070s, and 2090s) will experience different contraction and expansion movements relative to the current situation (Table 2; Fig. 5). Overall, compared with the suitable distribution range in the current situation, the suitable habitat area of S. salsa would only shrink by 0.85% under the SSP245 scenario in the 2050s; later, it would continue to expand with a maximum expansion area of approximately 16.11% under the SSP585 scenario in the 2050s (Fig. 6). Under the SSP245 and SSP585 scenarios, the suitable habitat area of S. europaea would be reduced by 9.00% and 9.18% in the 2050s, respectively (Fig. 7). The suitable habitat area for H. glomeratus would shrink completely under the two future climate scenarios (Fig. 8). Specifically, under the SSP585 scenario, the maximum contraction of the suitable habitat for H. glomeratus in the 2050s would be nearly 15.30%. The suitable habitat area of H. strobilaceum would be reduced by 10.72% in the 2050s and 4.01% in the 2090s (Fig. 9). On the other hand, the suitable habitat area of H. caspica would shrink by 1.29% and 3.82% in the 2070s and 2090s under the SSP245 scenario, respectively, and reduce by 4.66% in the 2070s under the SSP585 scenario (Fig. 10). Under the future climate scenarios, complete expansion would be observed in the suitable habitat area of K. foliatum, with the maximum expansion of the suitable habitat being nearly 6.27% in the 2050s (Fig. 11).

In the future (2050s, 2070s, and 2090s), the distribution areas of the suitable habitats for the six halophytic plant species under the SSP245 and SSP585 scenarios would gain or lose to varying degrees compared with the current conditions (Figs. 6-11). For S. salsa, the gain in the suitable habitat area would be 7.23×10⁴-14.24×10⁴ km² in eastern Inner Mongolia, Xinjiang, and Gansu, and the loss in the suitable habitat area would be 4.86×10⁴-7.69×10⁴ km², mainly in Inner Mongolia. It is predicted that for S. europaea, the gain in the suitable habitat area would be 4.38×10⁴-8.76×10⁴ km² in the eastern part of western and easternmost regions of Inner Mongolia and the Turpan Basin of Xinjiang, whereas the loss in the suitable habitat area would be 3.15×10⁴-6.96×10⁴ km², mainly in northern Xinjiang, central-eastern part of Inner Mongolia, and Qinghai. For H. strobilaceum, the suitable habitat area would greatly increase in Xinjiang and sporadically increase in Gansu and Inner Mongolia, with the gain of 3.52×10⁴-6.51×10⁴ km², whereas the loss in the suitable habitat area would be 5.33×10⁴-7.93×10⁴ km² in northern Xinjiang and central-eastern part of Inner Mongolia. For H. glomeratus, the suitable habitat area would sporadically increase in southern Xinjiang and Qinghai, with the gain of 2.84×10⁴-6.03×10⁴ km², whereas the loss in the suitable habitat area would be 11.94×10⁴-18.30×10⁴ km² in eastern Xinjiang and central Inner Mongolia. For H. caspica, the gain in the suitable habitat area at high latitudes would be 9.40×10⁴-14.25×10⁴ km², whereas the loss in the suitable habitat area would be 11.37×10⁴-17.15×10⁴ km² at low latitudes. The suitable habitat range of K. foliatum would gain eastward and northward along the distribution boundary of the current habitats, with the gain area of 12.12×10⁴-14.54×10⁴ km² and the loss area of 8.93×10⁴-14.36×10⁴ km².

At present, the centroid of the suitable habitat for S. salsa is located in the Alagxa Right Banner of Inner Mongolia (Fig. 4a). In the future, the centroid of the suitable habitat for S. salsa would be projected to shift to high latitudes under the SSP585 scenario. At present, the centroid of the suitable habitat for S. europaea is located in the southeast of Yizhou District, Hami City, Xinjiang (Fig. 4b). In the future, it would generally shift to the eastern region. At present, the centroid of the suitable habitat for H. strobilaceum is located in the western part of Toksun County, Xinjiang (Fig. 4c), and it would be generally shifted to the western part and high latitudes in the future. At present, the centroid of the suitable habitat for H. glomeratus is located in the northeast of Yuli County, Xinjiang (Fig. 4d), and it would shift to the east and high latitudes in the future under the two climate scenarios. At present, the centroid of the suitable habitat for H. caspica is located in the northern part of Ruoqiang County, Xinjiang (Fig. 4e), and in the future, this centroid will generally shift to high latitudes in the northern region relative to its present location. The centroid of the suitable habitat for K. foliatum is currently located in the central part of Guazhou County at the western end of Gansu (Fig. 4f). In the future, it would shift and alternately migrate along the east-west direction and generally be in Dunhuang City, Gansu under the SSP245 scenario.

In this study, the suitable habitats of S. salsa, S. europaea, H. glomeratus, H. strobilaceum, K. foliatum, and H. caspica in the arid areas of Northwest China were predicted in the future under two climate scenarios. The predicted AUC values for all plant species using the MaxEnt model were higher than 0.80, which met the evaluation criteria and had high accuracy (Phillips et al., 2006). In our study, the geographical occurrence data of the halophytic plant species were processed using the data processing strategy proposed by previous studies to prevent the overfitting of the model (Phillips and Dudík, 2008; Kramer-Schadt et al., 2013; Merow et al., 2013; Warren et al., 2014). The ENMTools software was used to optimize the species occurrence records and reduce the possibility of overfitting the model due to data problems. The MaxEnt model has demonstrated great accuracy and superiority over alternative approaches, even when available species occurrence data are scarce (Hernandez et al., 2006; Gogol-Prokurat, 2011). For instance, Gogol-Prokurat (2011) employed the MaxEnt model for predicting the species distribution to forecast the habitat suitability of uncommon plant species on a small geographical scale in El Dorado County, California, USA. Zhang and Zhao (2021) employed the MaxEnt model to predict biodiversity hotspots for rare and endangered plant species in Northwest China, providing scientific insights for habitat restoration and reconstruction efforts. Huang et al. (2019) used the MaxEnt model to predict the suitable habitat ranges of five plant species belonging to the genus of Zingiber in China; their results indicated that under climate change, the suitable habitats for most Zingiber species will significantly decrease, highlighting the need for the conservation of these plant resources. Therefore, the MaxEnt model is suitable for analyzing the current and future habitat distributions of the studied six halophytic plant species, namely, S. salsa, S. europaea, H. glomeratus, H. strobilaceum, K. foliatum, and H. caspica, in saline environments.

Pearson's correlation analysis and Jackknife test method were used to identify the primary environmental variables that impact the distribution of halophytic plant species (Elith et al., 2011; Fourcade et al., 2018). Finally, 16 environmental variables were selected for use in the MaxEnt model, and the contribution rates (Table S2) and suitable habitat ranges (Table S3) were obtained. The results revealed that climate was the primary environmental factor influencing the distribution of halophytic plant species compared with soil factors (Li et al., 2022). The contribution of three precipitation variables (Bio15, Bio18, and precipitation of the coldest quarter (Bio19)) to the distribution of six halophytic plant species was 40.98%. Previous studies have highlighted the significant influence of precipitation on the distribution patterns of plant species (Abolmaali et al., 2017; Khajoei Nasab et al., 2020; Liu et al., 2021). Precipitation affects the distribution patterns of plant species through its impact on seed germination, seedling development and survival, and phenology (Quevedo-Robledo et al., 2010; Mathias and Chesson, 2013). Future climate scenarios, including changes in precipitation, will influence the variations in different plant species in terms of positive or negative responses. For halophytes growing in saline-alkaline land in arid and semi-arid areas, if precipitation decreases in the future, the distribution pattern and suitable habitat area of plant species will significantly change (Song et al., 2023). For example, the distribution pattern of plant species will shift to higher latitudes and suitable habitat area will face loss and fragmentation, as the decrease of precipitation.

The contribution of temperature variables to the distribution of six halophytic plant species was 14.88% (Table S2). Of them, the contribution of Bio10 to the distribution of K. foliatum, H. strobilaceum, and S. salsa was relatively high among the temperature variables. Qu et al. (2008) reported that the germination rate of H. strobilaceum was higher at high temperatures than at low temperatures. Wang et al. (2019) discovered that the best germination temperature for S. glauca ranges from 10.00°C to 20.00°C. Different plant species exhibit different reponses to temperature; therefore, global warming may increase the distribution range of S. salsa, K. foliatum, and other halophytic plant species. According to recent research, temperature may significantly affect the potential distribution of halophytic plant species (Ardestani et al., 2015; Ma and Sun, 2018; Zhao et al., 2021). Precipitation significantly influences plant diversity in arid and semi-arid areas (Gessner et al., 2013; Li et al., 2013), followed by heat conditions in the growing season. Therefore, changes in precipitation and temperature exert a notable influence on the distribution of halophytic plant species, affecting plant development and seedling survival (Poorter, 2000).

In addition to climate variables, soil is another key environmental factor that influences the changes of suitable habitat and distribution of halophytic plant species. Alterations in the soil's physical and chemical characteristics, in addition to changes in the composition of bedrock, can affect the distribution of halophytic plant species (Piedallu et al., 2011). The tendency for soil salinization can be accelerated by increases in temperature and sea level, decrease in precipitation, and ineffective irrigation management (Eswar et al., 2021). S. salsa, K. foliatum, H. capsica, and S. europaea were greatly affected by soil factors, including Na. Excessive salt content in the soil reduces plant water absorption, disrupts metabolism, and inhibits plant growth (van Zelm et al., 2020). S. salsa exhibits moderate tolerance to salt stress (Wang et al., 2019); however, K. foliatum and H. caspica show strong tolerance to salt stress. The combination of drought and salinity can hinder plant growth (Lu et al., 2021a). In this study, the contribution rate of available nitrogen (AN) to H. glomeratus was the highest with the suitable range of 41.11-109.77 mg/kg, and that of cation exchange capacity (CEC_SOIL) to H. strobilaceum was the highest with the suitable range of 1.00-1.69 cmol/kg. This is consistent with the study by Nilhan et al. (2008) on the impact of CEC_SOIL on the distribution of H. strobilaceum. The content of soil organic matter significantly influences the occurrence and distribution of halophytic plant species. Recent research has indicated that soil variables significantly influence the distribution of halophytic plant species (Yilmaz et al., 2017; Khajoei Nasab et al., 2020).

Changes in precipitation, temperature, and other components of human-induced climate change alter the species distribution on a global or local level (Pecl et al., 2017). Species distribution ranges would expand, contract, or remain unchanged due to global warming (Petitpierre et al., 2016). In this study, various halophytic plant species exhibited distinct ecological habits in response to climate change. As the impact of climate change worsen in the future, the suitable distribution range of S. salsa will expand with global warming and there is a trend of westward and nearby expansion (Fig. 6). This may be related to the fact that the developmental state of S. salsa is directly impacted by water and salt (An et al., 2019). The suitable habitat area of S. europaea would significantly decrease in the 2050s under two climate scenarios. However, the habitat loss area would gradually decrease and become fragmented, which is obvious in northern Xinjiang, Inner Mongolia, and Qinghai (Fig. 7). S. europaea was observed to be relatively highly affected by precipitation and soil variables (including Na and pH). Climate change impacts plant water supplies and hinders seed germination (Shi et al., 2021). The suitable habitat area of H. glomeratus would shrink to varying degrees at any time and under any climate scenario in the future, and the shrinking trends would occur in Inner Mongolia and Xinjiang (Fig. 8). The shrinkage of the suitable habitat area for H. glomeratus in the future may be affected by the gradual decrease in precipitation and increase in temperature (Korell et al., 2021). One of the primary influential factors in habitat fragmentation is thought to be climate change (Yi et al., 2016; Qin et al., 2017). For example, the suitable habitat range of the globally cultivated Xanthium italicum Moretti will shrink in the future due to climate change (Zhang et al., 2021). Although the current study is not directly related to habitat fragmentation, shifts in species distribution by climate change are very likely to result in habitat fragmentation. Habitat fragmentation can lead to a reduction in local plant species populations and the flow of genes from other populations (Wu et al., 2014). As a result, adaptability, genetic variation, and diplomacy will be reduced (Wang et al., 1998). The suitable habitat area of H. strobilaceum will exhibit complete expansion under the SSP585 scenario in the future (Fig. 9); currently, this species also faces the risk of habitat loss in northern Xinjiang and Inner Mongolia. This is consistent with the results of Bezeng et al. (2017) regarding the expected effects of climate change on 162 non-indigenous plant species in South Africa. Bezeng et al. (2017) indicated the likely expansion of species ranges in specific areas by the 2070s. The change in the distribution area of H. caspica would exhibit some fluctuations. Contraction and expansion of suitable habitat area are equally possible, but the scope of change is minimal. An obvious contraction would occur in the central and eastern Xinjiang and in the intersecting area of Ningxia and eastern Inner Mongolia (Fig. 10). Climate change can lead to insufficient water and excessive salt content in the soil (Hassani et al., 2021). Although the suitable habitat area of K. foliatum shows a trend toward complete expansion in the future, this species will face the risk of habitat loss in the eastern part of Inner Mongolia and in the Taklimakan Desert of Xinjiang due to changes of precipitation variables (Fig. 11). Increased soil salt content leads to decreased seed germination rate, and temperature, precipitation, soil salinity, and their interactions further affect seed germination (Zhang et al., 2012). Climate change may be good or bad for the living conditions of some halophytic plant species (Lososová et al., 2018). For example, the distribution of six endangered saproxylics was predicted by integrating diffusion, land use change, and climate change (Rocca and Milanesi, 2020); the findings indicated that the distributions of four of the six halophytic plant species in our study were positively impacted by climate change.

It is very important to protect the shrinking habitat area of halophytes in the arid areas of Northwest China, which will be further exacerbated by global warming and the ensuing drought in the future (Qiu et al., 2017). To adapt to climate change, many halophytes would shift to more favorable habitats or face extinction; therefore, climate has an important impact on the distribution patterns of halophytic plant species (Corlett and Westcott, 2013). The spatiotemporal dynamics of widely distributed plant species should be detected and studied in detail (Collen et al., 2016). The impact of habitat fragmentation on the adaptation of halophytes to future climate change should also be further studied in detail.

Climate change impacts the migration of the suitable habitats of halophytic plant species (Richman et al., 2020). In the 2050s, the future centroid of the suitable habitat for K. foliatum would shift to the southwest after eastward migration under the SSP245 scenario, mainly due to the influences of temperature and precipitation (Corlett and Westcott, 2013). Although the centroids of the suitable habitats of S. salsa, H. capsica, H. glomeratus, and H. strobilaceum would shift to high latitudes with global warming, the migration distance among plant species would be different; the maximum shift distance of the centroid for each plant species would be 43.35-132.89 km, with variances for widely and narrowly distributed species. Plant species with restricted distributions are more susceptible to climate change because of their limited ecological adaptability (Zhang et al., 2015; Ma and Sun, 2018). For example, S. salsa and H. capsica are more widely distributed, and their centroid migration is smaller. Plant species that are more adaptable to climate change will have wider distribution ranges (Hu et al., 2015).

Climate is a crucial ecological component. Plant species adapt to climate change by altering their geographical distribution, such as by moving to higher latitudes or altitudes to find favorable environmental conditions (Chen et al., 2011; Lehikoinen and Virkkala, 2016; Qin et al., 2017). The centroid displacement of the suitable habitats for the six halophytic plant species in the future under the SSP245 and SSP585 scenarios indicated a relatively weak migration in case of the moderate SSP245 scenario, which is the same for several plant species spread over the western and central regions of China in the future (Chen et al., 2019; Sun et al., 2020). Under climate change, halophytic plant species will migrate to more suitable climatic zones, adjust their phenological or physiological adaptations to the new climate conditions, or become extinct in response to climate change (Vitt et al., 2010). According to previous study, certain plant species have apparently started to migrate, and most of these species cannot keep up with the impact of climate change (Corlett and Westcott, 2013). For example, migration of halophytic plant species for suitable habitats is strongly correlated with precipitation, temperature, soil, and other factors. It may mean that in case of long-term or dramatic climate change, the suitable habitats of halophytic plant species will be seriously affected by climate change (Pigot et al., 2023). This impact may be influenced by the complexity of topography and suitable habitat conditions for halophytic plant species, resulting in different degrees of centroid migration.