Threatened Status Assessment of Multiple Grassland Ecosystems and Conservation Strategies in the Xilin River Basin, NE China

The Red List of Ecosystems, proposed by the International Union for Conservation of Nature can determine the status of ecosystems for biodiversity conservation. In this study, we applied the Red List of Ecosystems Categories and Criteria 2.0 with its four major criteria (A, B, C, and D) to assess twelve dominant ecosystems in the Xilin River Basin, a representative grassland-dominating area in China. We employed Geographical Information Systems and remote sensing to process the obtained satellite products from the years 2000 to 2015, and generated indicators for biological processes and degradation of environment with boreal ecosystem productivity simulator. The results show that all twelve ecosystems in the Xilin River Basin confront varying threats: Artemisia frigida grassland and Festuca ovina grassland face the highest risk of collapse, sharing an endangered status; Filifolium sibiricum meadow grassland and Leymus chinensis grassland have a least concern status, while the remaining eight ecosystems display a vulnerable status. This study overcomes the limits of data deficiency by introducing the boreal ecosystem productivity simulator to simulate biological processes and the plant–environment interaction. It sheds light on further application of the Red List of Ecosystems, and bridges the research gap and promote local ecosystems conservation in China.


Introduction
Extinction risk is an important factor considered in the process of setting priorities for biodiversity conservation and further sustainable development. The International Union for Conservation of Nature (IUCN) is regarded as a pioneer in extinction risk assessment and biodiversity conservation. Although the IUCN Red List of Threatened Species (RLTS) has been widely accepted and used to identify the threat status of species, it still has potential limitations in terms of biodiversity assessment at a broader scale [1]. To complement the RLTS, the IUCN held a workshop to develop an ecosystem-centered assessment in 2008. In 2014, the IUCN Council adopted the Red List of Ecosystems (RLE) Categories and Criteria 2.0 as an official global standard for assessing the risks to ecosystems [1]. Ecosystem assessment based on the RLE can provide multiple indicators of biotic diversity, while individual species assessment is limited in its ability to represent biodiversity as a whole. Additionally, ecosystem assessment considers the decline in the extent and status of habitats and ecosystems [1,2]. Furthermore, ecosystem assessment can be applied from the regional to a global level with the incorporation of geographical information systems (GIS) and remote sensing (RS) data, while species assessment is more Sustainability 2020, 12, 1084 3 of 17

The Study Area
The XRB, encompasses Xilin Hot City and Keshikten Banner ( Figure 1A), Inner Mongolia, within a northern, semi-arid, and farming-pastoral ecotone of China. The XRB is a representative grassland-dominant and biodiversity-rich area in China. In the basin, grassland accounts for 89% of the total study area and contains 629 spermatophyte species in 291 genera from 74 families including a large number of Eurasian initial species [24]. The Basin covers an area of 33,401 km 2 and ranges from 43 • 26 ~44 • 39 N to 115 • 32 ~117 • 12 E [25], with the elevation decreasing from southeast to northwest [26]. Total of 80% of the XRB is covered by grassland, with Leymus chinensis grassland and Stipa capillata grassland as the dominant ecosystems [27]. The data at the Xilin Hot Meteorological Station shows that the mean annual precipitation is 272 mm, 80-86% of which occurring from May to September, showing a semi-arid continental climate. The mean annual temperature is 2.59 • C, with an average of 19.19 • C in the summer and −16.67 • C in the winter [28]. Spatially different soil types, including chernozem soil, dark chestnut soil, and light chestnut soil, stretch from southeast to northwest [29][30][31].
Sustainability 2020, 12, 1084 3 of 18 to facilitate the identification of conservation priorities and the improvement of monitoring programs in the XRB.

The Study Area
The XRB, encompasses Xilin Hot City and Keshikten Banner ( Figure 1A), Inner Mongolia, within a northern, semi-arid, and farming-pastoral ecotone of China. The XRB is a representative grasslanddominant and biodiversity-rich area in China. In the basin, grassland accounts for 89% of the total study area and contains 629 spermatophyte species in 291 genera from 74 families including a large number of Eurasian initial species [24]. The Basin covers an area of 33,401 km 2 and ranges from 43°26′~44°39′ N to 115°32′~117°12′ E [25], with the elevation decreasing from southeast to northwest [26]. Total of 80% of the XRB is covered by grassland, with Leymus chinensis grassland and Stipa capillata grassland as the dominant ecosystems [27]. The data at the Xilin Hot Meteorological Station shows that the mean annual precipitation is 272 mm, 80-86% of which occurring from May to September, showing a semi-arid continental climate. The mean annual temperature is 2.59 °C, with an average of 19.19 °C in the summer and −16.67 °C in the winter [28]. Spatially different soil types, including chernozem soil, dark chestnut soil, and light chestnut soil, stretch from southeast to northwest [29][30][31].

Ecosystem Classification Data
Land cover data and vegetation data were obtained from the Ministry of Ecology and Environment Centre for Satellite Application on Ecology and Environment (China) (http://www.secmep.cn). Thirteen land cover types were classified based on the land cover classification system (LCCS) defined by the Food and Agriculture Organization of the United Nations (FAO). To determine the precision of the RS interpretation, the classification results were assessed via Kappa coefficients. The Kappa coefficients for 2000 and 2015 are 86.03% and 83.55%.
In this study, we assessed the ecosystem at the genus level. According to the Database for Ecosystems and Ecosystem Services Zoning in China, each ecosystem genus is determined by constructive species (co-constructive species) [32]. We named ecosystems with its constructive species: Ostryopsis davidiana shrub, Sanguisorba officinalis meadow, Stipa capillata grassland, Artemisia frigida grassland, Cleistogenes squarrosa grassland, Leymus chinensis grassland, Achnatherum splendens grassland, sandy pioneer plant communities (grassland), Festuca ovina grassland, Leymus chinensis meadow grassland, Stipa capillata meadow grassland, and Spiraea salicifolia shrub ( Figure 1B). Twelve ecosystems were grouped into nine ecosystem families and three ecosystem orders based on the IUCN Habitats Classification Scheme 3.0 in combination with Database for Ecosystems and Ecosystem Services Zoning in China [32]. These twelve dominant ecosystems, occupying more than 70% of the area were selected as the assessment targets.

Other Data
Leaf area index (LAI), normalized difference vegetation index (NDVI), meteorological, and soil data were used in the BEPS to develop the indicators for evaluating ecosystem status. We utilized the MOD13A1 and MOD15A2 16-day NDVI product and LAI product at a 500-m spatial resolution in 2000 and 2015, respectively (derived from EOS/MODIS of NASA) [33]. We applied the MODIS Reprojection Tools (the National Center for Supercomputing Applications, Chicago, US) to covert downloaded MODIS-NDVI and MODIS-LAI from HDF to TIFF and re-project output data to WGS84. Monthly MODIS-NDVI and monthly MODIS-LAI were obtained from 16-day NDVI product and 16-day LAI product based on maximum value composite (MVC) in ArcGIS (Esri, California, US). The XRB zoning map was used to extract the monthly NDVI and LAI grids in the region from 2000 to 2015. The meteorological data, including the average monthly temperature, monthly precipitation, monthly net radiation, and monthly mean relative humidity during 2000~2015, were collected from the Chinese Meteorological Data Sharing Service Network [34]. Data on the available water-holding capacity of soil were compiled in a GIS soil database at the geographical information monitoring cloud platform [35].

Ecosystem Assessment System
According to the IUCN criteria, ecosystems were classified into eight categories. Initially, all ecosystems are considered not evaluated (NE). Ecosystems should be evaluated using all criteria. If the data for all criteria are unavailable, the ecosystem should be classified as data deficient (DD). Once data are available, the status of an ecosystem is assigned to one of the remaining six categories based on the criteria and threshold: least concern (LC), near threatened (NT), vulnerable (VU), endangered (EN), critically endangered (CR), and collapsed (CO). The assessment was conducted at the ecosystem level instead of the single-species level, which is consistent with other the IUCN RLE-based studies.

•
The absolute rate of decline According to the Guidelines for the Application of IUCN Red List of Ecosystems Categories and Criteria, the time frames of assessment (under criteria A, C, and D) include the past 50 years, any 50 years and since 1750 [36]. Because of the data deficiency, assessments of changes in the past were not conducted. Assessment of future changes was based on the predictions of changes over any 50 years. Two scenarios including the proportional rate of decline (PRD) and the absolute rate of decline (ARD) were proposed to predict the future changes. Here we assume that variables decrease linearly every year. This model might introduce uncertainty, more efforts are needed to generate a more realistic model. We made predictions of changes for any 50 years (2000-2050) based on ARD over 15 years. A 15-year interval is common in the IUCN RLE-based grassland studies [2]. It is consistent with official guidelines for the RLE as well. The Guidelines for the Application of IUCN Red List of Ecosystems Categories and Criteria 2.0 provides an example study using 1986-2001 to first estimate an observed rate of change over 15 years, and then extrapolating projected losses to 2036 [36]. Given the above-mentioned reasons, we consider the time interval here is appropriate. The change in the ecosystems from 2000 to 2050 can be predicted from Equation (1): where A t0 is the value at year t 0 and A t1 is the value at year t 1 . In the present study, the temporal scale (t 1 − t 0 ) is 15 years from 2000 to 2015. R is a proportion of change (%).

• Criterion A: reduction in geographic distribution
Criterion A is to identify ecosystems that are undergoing a decline in areas that influences the risk of collapse [5]. In this study, we evaluated the threat status of ecosystems under subcriterion A2 (reduction over any 50 years) by calculating the area they occupied. Future change assessment was based on the prediction of changes of geographical distribution from 2000 to 2050 calculated by the equation (1).

•
Criterion B: restricted geographic distribution The primary role of criterion B is to identify ecosystems whose distributions are restricted to the extent, facing the risk of collapse under threatening events or processes [37]. Criterion B was used to assess restricted distribution in terms of the extent of occurrence (EOO) (B1), the area of occupancy (AOO) (B2), and the number of locations (served as B3 here). The EOO was estimated using a minimum convex polygon enclosing all the sites within an ecosystem type. The AOO was calculated by counting the number of 10 × 10 km grains within the EOO covered by an ecosystem.

•
Criterion C: environmental degradation Criterion C evaluates the risk of ecosystem collapse caused by the degradation of the abiotic environment [38]. It was used to assess environmental degradation over the past 50 years (C1), any 50 years (C2), and since 1750 (C3) based on the changes in abiotic variables. In the analysis, soil water deficit (SWD) and soil carbon (SC) were selected as environmental indicators to assess environmental degradation. SWD and SC in 2050 were calculated by the Equation (1). Regarding the guidelines, the key to assessing criteria C and D is relative severity. It is because relative severity is essential for comparing risks among ecosystems suffering from different degradation [36]. In this study, the values of parameters in 2000 are regarded as the initial values and collapse value is assumed to be zero. The relative severity of decline is calculated as where P t0 represents the parameter in 2000 and P t1 represents the parameter in 2050. Criterion D considers the extent of the ecosystem that is subjected to the disruption of biotic processes and the severity of disruption [39]. Here, the dimidiate pixel model and BEPS were applied to simulate biological processes and construct biological indicators. Vegetation fractional coverage (VFC), vegetation evaporation (EV), net primary production (NPP), belowground net primary production (BNPP), aboveground net primary production (ANPP), gross primary production (GPP), and net ecosystem production (NEP) are selected to represent the biological process. The change of these seven indicators from 2000 to 2050 is calculated by the Equation (1). The relative severity of decline was calculated by the Equation (2) as well. In this contribution, we only focus on the dynamics of vegetation. We realize that ecosystems have multiple components and fauna is important as well. Yet it is still challenging not only for our study but for other studies [21]. Fauna is highly movable and it is difficult to track them every year and make predictions for several decades. Besides, interactions between fauna and flora are complicated, and it might cause more uncertainties for assessment if we could not tackle this issue properly. More importantly, since quantification and assessment of fauna indicators for the RLE have not had consistency so far, it is hard to assess them [36]. More efforts will be needed to improve the assessment of biotic processes and interactions.

•
Determining the threat level For criteria C and D, we calculated the relative severity of the decline. Then we categorized relative severity (≥30%) into three degrees: ≥80% (d1), ≥50% (d2), and ≥30% (d3). Different relative severity degrees were combined with a proportion of area they occupied to assign threat levels for twelve ecosystems. Table 1 is an overview of sub-criteria and thresholds adopted in this study. Following the precautionary principle, the overall risk status of the ecosystem will be the highest risk category obtained by any of the assessed criteria.

Boreal Ecosystem Productivity Simulator (BEPS)
Process-based biogeochemical models can integrate the input data, automate the modeling, and produce the output of ecological parameters. It has been a useful tool for revealing mechanisms of biological process and plant-environment interaction. BEPS is a more reliable process model than others in terms of availability and quality of required data and accuracy of simulation results. Accuracy of BEPS simulation is estimated 60% for single pixel (1 km 2 ) and 75% for 3 × 3 pixels (9 km 2 ) [40]. In this study, nine indicators were developed based on the BEPS to assess environmental degradation and the decline of biotic processes (Table 2). More details about BPES can be found in the reference paper.
To evaluate the accuracy of model results, we compared the NPPs of Stipa capillata grassland and Festuca ovina grassland obtained from ground measurements with those obtained from the BEPS. In 2015, for Stipa capillata grassland, the R 2 value of the relationship between observed and modeled NPP was 0.725, whereas that of Festuca ovina grassland was 0.603. According to the previous studies [41][42][43][44], Sustainability 2020, 12, 1084 7 of 17 the modeled results are appropriate for use in an ecosystem assessment system. An overview of the conceptual framework is shown in Figure 2. Table 2. Outputs of boreal ecosystem productivity simulator (BEPS) selected as parameters for criteria C and D.

Inputs Outputs Description Criterion
Remote sensing data, soil data, meteorological data SWD SC Soil health and quality [33,34]. C VFC EV NPP BNPP ANPP GPP NEP VFC reflects soil carbon, sequestration and water conservation [34]. Primary productivity is the foundation that sustains the abiotic processes in ecosystems and it can be expressed by ANPP, BNPP, GPP, NPP, and NEP [36]. D Sustainability 2020, 12, 1084 7 of 18 observed and modeled NPP was 0.725, whereas that of Festuca ovina grassland was 0.603. According to the previous studies [41][42][43][44], the modeled results are appropriate for use in an ecosystem assessment system. An overview of the conceptual framework is shown in Figure 2. Table 2. Outputs of boreal ecosystem productivity simulator (BEPS) selected as parameters for criteria C and D.

Inputs Outputs Description Criterion
Remote sensing data, soil data, meteorological data SWD SC Soil health and quality [33,34]. C VFC EV NPP BNPP ANPP GPP NEP VFC reflects soil carbon, sequestration and water conservation [34]. Primary productivity is the foundation that sustains the abiotic processes in ecosystems and it can be expressed by ANPP, BNPP, GPP, NPP, and NEP [36].
D Figure 2. The conceptual framework showing inputs/outputs data of BEPS and spatial data serve as inputs data for the assessment of ecosystems.

Criterion A, Reduction in Geographic Distribution
Reduction from 2000 to 2050 (A2) From 2000 to 2015, the total area had decreased by 0.5%. Stipa capillata grassland occupied the greatest area, and from 2000 to 2015, it experienced the most noticeable decrease of 93.4 km 2 . Artemisia frigida grassland accounted for the smallest proportion of the total area. From 2000 to 2050, Stipa capillata grassland, Achnatherum splendens grassland, and sand pioneer plant community ecosystems were simulated to decrease by 41.22%, 40.01%, and 42.54%, respectively, while the remaining ecosystems experienced slight increases (Figure 3). According to sub-criterion A2 (decrease > 30%), Stipa capillata grassland, Achnatherum splendens grassland, and sand pioneer plant community ecosystems were assigned to the VU category, whereas the other nine ecosystems were assigned to

Criterion A, Reduction in Geographic Distribution
Reduction from 2000 to 2050 (A2) From 2000 to 2015, the total area had decreased by 0.5%. Stipa capillata grassland occupied the greatest area, and from 2000 to 2015, it experienced the most noticeable decrease of 93.4 km 2 . Artemisia frigida grassland accounted for the smallest proportion of the total area. From 2000 to 2050, Stipa capillata grassland, Achnatherum splendens grassland, and sand pioneer plant community ecosystems were simulated to decrease by 41.22%, 40.01%, and 42.54%, respectively, while the remaining ecosystems experienced slight increases (Figure 3). According to sub-criterion A2 (decrease > 30%), Stipa capillata grassland, Achnatherum splendens grassland, and sand pioneer plant community ecosystems were assigned to the VU category, whereas the other nine ecosystems were assigned to LC. Because of the lack of data necessary to evaluate the geographic changes over the past 50 years and since 1750, the status of all twelve ecosystems under criteria A1 and A3 are listed as DD. LC. Because of the lack of data necessary to evaluate the geographic changes over the past 50 years and since 1750, the status of all twelve ecosystems under criteria A1 and A3 are listed as DD.

EOO (B1)
Stipa capillata grassland was the most widely distributed ecosystem type, covering an area of 31,154 km 2 , whereas Festuca ovina grassland was the most limited type (Table 3, Figure 4). Artemisia frigida grassland, Cleistogenes squarrosa grassland, Ostryopsis davidina shrub, Sanguisorba officinalis meadow grassland, sand pioneer community ecosystem, and Festuca ovina grassland each occupied less than 10,000 km 2 ; and according to sub-criterion B1, they all have VU status. The remaining ecosystems were assigned into LC.

EOO (B1)
Stipa capillata grassland was the most widely distributed ecosystem type, covering an area of 31,154 km 2 , whereas Festuca ovina grassland was the most limited type (Table 3, Figure 4). Artemisia frigida grassland, Cleistogenes squarrosa grassland, Ostryopsis davidina shrub, Sanguisorba officinalis meadow grassland, sand pioneer community ecosystem, and Festuca ovina grassland each occupied less than 10,000 km 2 ; and according to sub-criterion B1, they all have VU status. The remaining ecosystems were assigned into LC.

AOO (B2)
The area occupied by the ecosystems was at 10 × 10 km grain size, and those occupying less than 1 km 2 were excluded as Figure 5 shows. Analogous to the geographic distributions, Stipa capillata grassland accounted for the largest number, with 178 grid cells, whereas Festuca ovina grassland accounted for the smallest number, with only nine grid cells. According to the evaluation sub-criterion B2, Festuca ovina grassland, which occupied fewer than 20 grid cells, belongs to the EN category. Artemisia frigida grassland, Cleistogenes squarrosa grassland, Leymus chinensis meadow grassland, Sanguisorba officinalis meadow grassland, and sand pioneer community ecosystem occupied less than 50 grid cells and were assigned to the VU category. The remaining ecosystems belonged to the LC category.

AOO (B2)
The area occupied by the ecosystems was at 10 × 10 km grain size, and those occupying less than 1 km 2 were excluded as Figure 5 shows. Analogous to the geographic distributions, Stipa capillata grassland accounted for the largest number, with 178 grid cells, whereas Festuca ovina grassland accounted for the smallest number, with only nine grid cells. According to the evaluation subcriterion B2, Festuca ovina grassland, which occupied fewer than 20 grid cells, belongs to the EN category. Artemisia frigida grassland, Cleistogenes squarrosa grassland, Leymus chinensis meadow grassland, Sanguisorba officinalis meadow grassland, and sand pioneer community ecosystem occupied less than 50 grid cells and were assigned to the VU category. The remaining ecosystems belonged to the LC category. According to the sub-criterion B3, easily affected locations of each ecosystem were counted. Festuca ovina grassland had less than five locations, and it belonged to the EN category. Artemisia frigida grassland, Sanguisorba officinalis meadow grassland, and sand pioneer community ecosystem

Number of Locations (B3)
According to the sub-criterion B3, easily affected locations of each ecosystem were counted. Festuca ovina grassland had less than five locations, and it belonged to the EN category. Artemisia frigida grassland, Sanguisorba officinalis meadow grassland, and sand pioneer community ecosystem occupied less than ten locations; thus, these ecosystems belonged to the VU category. The remaining ecosystems belonged to the LC category.

Criterion C, Environmental Degradation
Environmental Degradation from 2000 to 2050 (C2) From 2000 to 2050, none of the ecosystems showed noticeable declines in SWD and SC, and areas with decreasing trends in environment quality accounted for less than 1% of the total (Figures 6 and 7). According to sub-criterion C2, twelve ecosystems were assessed as LC. C1 and C3 are not assessable because of data deficiency.

Disruption of Biotic Process and Interactions from 2000 to 2050 (D2)
The biotic interactions and biological processes were quantitatively assessed by ANPP, BNPP, GPP, NPP, NEP, EV, and VFC. Total of 93.45% of Artemisia frigida grassland and 52.36% of Festuca ovina grassland were assigned into d2 and d1 respectively. Ostryopsis davidina shrub in d3 and Stipa capillata meadow grassland in d2 accounted for 81.15% and 35.44% respectively. According to sub-criterion D2, Artemisia frigida grassland and Festuca ovina grassland were at the EN level. Ostryopsis davidina shrub and Stipa capillata meadow grassland were at the VU level. D1 and D3 are not assessable because of data deficiency.

Assessment Result of Twelve Ecosystems
Based on RLE, we assigned the threat levels to the twelve ecosystems in the XRB. As Table 4 shows, Artemisia frigida grassland and Festuca ovina grassland faced EN-level threat; Filifolium sibiricum meadow grassland and Leymus chinensis grassland had LC status; and the remaining are at the VU level, which was the most common.
Different ecosystems faced different threats caused by various factors (Figure 8). Artemisia frigida grassland and Festuca ovina grassland faced the highest risks of collapse because of the declines in restricted distribution and environmental quality. Compared with Artemisia frigida grassland, Festuca ovina grassland experienced a more severe decline in restricted distribution. Achnatherum splendens grassland and Stipa capillata grassland were identified to be vulnerable resulting from the degradation in geographic distribution, whereas Leymus chinensis meadow grassland, Sanguisorba officinalis meadow, and Cleistogenes squarrosa grassland were identified as vulnerable because of the declines in restricted distribution and degradation. Ostryopsis davidina shrub and Stipa capillata meadow grassland suffered from the degradation because of environmental quality decline, whereas sand pioneer community ecosystem faced threat caused by declines in geographic distribution and restricted distribution. Thus, to conserve the different ecosystems, different measures should be taken. For example, for Artemisia frigida grassland and Festuca ovina grassland, measures on conservation of habitat can be taken. declines in restricted distribution and degradation. Ostryopsis davidina shrub and Stipa capillata meadow grassland suffered from the degradation because of environmental quality decline, whereas sand pioneer community ecosystem faced threat caused by declines in geographic distribution and restricted distribution. Thus, to conserve the different ecosystems, different measures should be taken. For example, for Artemisia frigida grassland and Festuca ovina grassland, measures on conservation of habitat can be taken.

Implications and Limitations
Twelve selected local ecosystem genera were evaluated by the IUCN RLE in combination with GIS, RS, and BEPS. According to , RLE-based ecosystem assessments should be judged by whether it achieves conservation and management ends, whether its limitations can be compensated by its advantages and benefit, and whether it performs better than alternative methods [45]. In this study, the threat status of each ecosystem was tracked and threat levels were assigned to each ecosystem systematically. It will promote its current and potential applications in legislation, policy, environmental management, and education in China. Besides, it can be more widely used and efficient than other current assessments in China. Considering these aspects, we think our ecosystem-centered study is appropriate. In addition, results are consistent with the previous studies, indicating that the status of multiple ecosystems in Inner Mongolia can be assessed by the RLE [5,[36][37][38][39]. It demonstrates the feasibility and possibility of applying the IUCN RLE in China. It also proves that data deficiency can be overcome to some extent by introducing the BEPS. BEPS as a virtual lab allows us to get biological indicators of vegetation and environment with high accuracy. It can be improved as knowledge, data availability, and quality improve [46]. BEPS can be adjusted and applied with higher accuracy in the future study.
However, there are potential limitations in this study. The persistence of biota within an ecosystem relies on biotic processes and interaction. These include predatory, species invasions, and trophic and pathogenic processes, et cetera. Diversity of organisms and process are important. Significant disruptions in process and interactions can even cause a collapse of the ecosystem [1,36]. However, in this study, our results only depend on the biological process and dynamics of vegetation. Other important elements in the ecosystem such as animal activities, trophic diversity, a spatial flux of organisms, interaction diversity are not taken into account. Though our assessment performs well, more effort in selecting biodiversity indicators and presenting complicity of the ecosystem is needed. Besides, the assessment still has uncertainties in spatial and functional symptoms [36]. Possible errors in mapping and classification can result in biased distribution. Because of the lack of data, we failed to calculate the past decline to distinguish directional change. For the future, study about reducing assessment errors and developing a more comprehensive model is necessary.

Future Predictions and Conservation Strategy for Local Ecosystems
In the XRB, Artemisia frigida grassland and Festuca ovina grassland face the highest threat level, EN, because of declines in biological processes and restricted distribution. If the trend persists, Artemisia frigida grassland and Festuca ovina grassland will be at the risk of collapse in the future. Other ecosystems with VU classification face survival challenges from geographic and restricted distribution. They would face higher levels of threats unless effective ecological management programs are implemented.
In the past decades of the XRB, the irregularity of patches and the degree of fragmentation have increased, with the area of dominant patches and patch connectivity decreasing [47,48]. Landscape alteration in the XRB is detrimental to ecosystem growth since it limits species immigration and emigration, which further affects the predictor-prey network and leads to the gene pool and biodiversity loss. Land use considered the main driver of landscape change should be managed properly. According to the previous studies, during the past decades, the area of grassland has decreased with city expansion, farmland reclamation, and mine exploration in Xilin Hot. Besides, mines in Xilin Hot are concentrated in grassland in the XRB, which increase local habitat fragmentation [49]. Control of the intensity of mining, allocation of mining spots, and management of the human landscape are important for conserving local ecosystems. Establishing the protected areas and land use management system is beneficial for protecting ecosystems against degradation. The ANPP of degraded ecosystems increased by 69% and vegetation cover increased by 45% after the areas were fenced [30].

Biodiversity Conservation Based on the IUCN RLE in China
Though the RLE can be used in the study area, there are some potential limitations in applying it widely. To improve the assessment and promote the application of the IUCN RLE in China, additional progress is needed. Except for variables mentioned in this study, a broad set of variables that are potentially useful for quantifying biotic processes and associated functional declines could be further considered. These could include changes in species richness, composition, and dominance; relative abundance of species functional types, guilds or alien species; measures of interaction diversity; measures of niche diversity and structural complexity [36]. For example, in China, by 2003, the number of invasive species reached 283, and invasive organisms can threaten local ecosystems through range expansion [50,51]. Thus, the system of ecosystem assessment could take a relative abundance of alien species into consideration to identify the major threats to local systems in regions suffering bio-invasion. Furthermore, there are inconsistencies between Chinese nature reserve classification and IUCN protected area categories in classification standards, management targets and functions [52]. To maximize the accuracy of comprehensive assessment, the adjustment of nature reserve classification to better correspond with IUCN protected area classification could be taken into accounts.
According to the China National Biodiversity Conservation Strategy and Action Plan (2011-2030), major challenges in biodiversity conservation include a lack of monitoring systems for biodiversity and limited awareness of biodiversity conservation [14]. The reasonable application of IUCN RLE can be of benefit to address these two major issues to some extent. The criteria are easily quantified and used, and they can directly reflect the status and changes of ecosystems. They will help establish a dynamic monitoring network for ecosystems in China [22]. Also, ecosystems whose services and functions are closely correlated with human beings are important for human wellbeing, and ecosystem assessment results can serve as a good educational tool to raise public awareness of biodiversity conservation [53]. The IUCN RLE can also contribute to the design of protected areas, smart city, natural resource management strategies, and frameworks for sustainable development [54].

Conclusions
In this study, we assessed the threat level of twelve dominant ecosystems on the base of IUCN RLE in the XRB by integrating the tools of geographical information systems, remote sensing, and boreal ecosystem productivity simulator. The integrated tools overcome the limits of data deficiency and increase the efficiency of the IUCN RLE to some extent. The results indicate that Artemisia frigida grassland and Festuca ovina grassland are facing EN-level threat; Filifolium sibiricum meadow grassland and Leymus chinensis grassland are at LC-level threatened status, and the remaining eight ecosystems are at VU-level threatened status. Artemisia frigida grassland and Festuca ovina grassland will be at the risk of collapse if this trend maintains. According to assessment results, conservation of habitat and land management are proposed for local biodiversity conservation. Although there are potential uncertainties and limitations, our preliminary results are consistent with previous studies. It demonstrates the applicational feasibility of the IUCN RLE in China and possibility of combining the IUCN RLE with the model. Our study will shed light on the future study, protection, and restoration of the XRB and further application of the IUCN RLE in China. For future studies, with the development of real-time tracking and computing, more aspects of the ecosystem will be explored and presented. The coverage of the IUCN RLE will be expanded and it will be possible to report biodiversity at the global level.