Assessing the degradation of grassland ecosystems based on the advanced local net production scaling method—The case of Inner Mongolia, China
Funding information: National Natural Science Foundation of China, Grant/Award Number: 41771460; National Key Research and Development Program of China, Grant/Award Number: 2017YFC0503803
Abstract
The gradual degradation of grasslands on a global scale goes hand-in-hand with significant challenges for agriculture and animal husbandry development. Numerous relevant policies and projects have been implemented to protect and restore Chinese ecosystems, but it is still unclear to what degree grassland ecosystems can be recovered. In view of this, we constructed an advanced local net production scaling (ALNS) method by replacing the classification method by self-organizing feature maps (SOFM) and tailoring the ideal state evaluation method in the LNS method. The ALNS method is used to analyze differences within grassland ecosystems, explore the ideal state of grassland ecosystems, and define degradation as the degree to which the actual state deviates from the ideal state, representing the degree to which grassland ecosystems can be recovered. It thereby quantifies and assesses the overall degradation of such systems in Inner Mongolia. Based on the results, more than 98.5% of the total grassland area failed to reach the ideal state, with the highest levels in the northeast with DN (degraded net primary productivity) values exceeding 200 gc/(m2·yr), followed by the midlands with DN values from 50 to 200 gc/(m2·yr) and the southwest with DN values between 0 and 150 gc/(m2·yr). The ALNS method can efficiently assess grassland ecosystem degradation and can be used to indicate the deviation degrees from ideal states, facilitating the development of protection and restoration programs for grassland ecosystems.
1 INTRODUCTION
Globally, grasslands account for about 26% of the total terrestrial area (Raney, Steinfeld, & Skoet, 2009). Grassland ecosystems perform multiple functions such as wind proofing and sand fixation, soil and water conservation, and biodiversity protection (Blair, Nippert, & Briggs, 2013; Hungate et al., 2017; Iravani et al., 2019). At the same time, they are the basis of livestock production, with a significant economic importance (Brand, 2009; Cao, Zhang, & Su, 2018; Zhen et al., 2010). In recent years, grassland degradation has become an important issue, in particular, against the background of a changing climate, and, currently, about 20% of the world's grasslands are degraded (Gao, Li, Xu, Wan, & Jiangcun, 2014; Odriozola, García-Baquero, Laskurain, & Aldezabal, 2014; Raney et al., 2009). As the country with the second-largest grassland area in the world (only followed by Australia), 47.7% of China is covered by grassland, and such areas account for 11.8% of the world total grassland area (Schweiger et al., 2015; Wang, Han, Cui, & Zhao, 2010; Zhang et al., 2018). As a result of climate change and human activities, grassland degradation increased from 10 to 90%, from the 1970s to the 2000s, and about 27.3% of natural grasslands are affected by desertification (Han et al., 2008; Liu et al., 2018). China's grassland ecosystem has experienced large-scale degradation and desertification, which have caused serious environmental problems and is affecting the Nation's food security (Akiyama & Kawamura, 2010; Li, Li, Guo, Liang, & Wang, 2016).
Inner Mongolia is the most degraded grassland region in China (Cao, Liu, Wei, Zheng, & Zhongqi, 2018; Su et al., 2017). The grassland ecosystems in Inner Mongolia represent an important ecological barrier in the north of the Country and play an important role in the social economy (Ouyang et al., 2018; Xiao, Xiao, Luo, Song, & Bi, 2019). Due to the influence of climate warming, along with prolonged droughts and overgrazing, the grassland in Inner Mongolia has suffered serious degradation, and approximately 90% have been degraded over a few decades (Akiyama & Kawamura, 2010; Shao, Chen, Zhang, & Huo, 2017), which led to myriads of environmental and economic problems, such as deterioration of biodiversity and ecosystem function and services and the aggravation of sandstorms, desertification, and poverty (Han et al., 2018a; Qi, Chen, Wan, & Ai, 2012; Shao & Dong, 2006). To mitigate grassland degradation and to promote the restoration of grassland ecosystems, a series of national grassland protection projects have been implemented in Inner Mongolia since 2000 (Jixia, Qibin, Jing, Depeng, & Quansheng, 2018; Li & Qin, 2014), for example, the Beijing–Tianjin Wind/Sand Source Control Program (BTWSSC), the Grazing Withdrawal Program, the Ecological Subsidy and Award System (ESAS), and the Natural Forest Protection Project (NFPP) (Dang, Li, Li, & Dou, 2018; Deng, Zhang, Cheng, Hu, & Chen, 2019; Feng, Tian, Yu, Yin, & Cao, 2019; Hua, Huang, & Li, 2019; Wang, Zhao, Fu, & Wei, 2019). In 2004, key ecological construction projects were first developed, and numerous studies have shown that the implementation of such projects has improved the ecological integrity of some areas of Inner Mongolia, although others are still being degraded (Cao, Liu, & Yu, 2018; Han, Song, Deng, & Xu, 2018; Yin, Hou, Langford, Bai, & Hou, 2019; Zhou & Zhao, 2017).
Numerous studies have focused on the degradation/restoration of grassland ecosystems in Inner Mongolia with the following objectives: (a) To illustrate the actual changes of grassland ecosystems. From the perspective of grassland ecosystem productivity, the net primary productivity (NPP) of the grassland ecosystem in Inner Mongolia showed a rough ‘U’ curve variation from 1988 to 2014, with a decrease from 1988 to 2004 and an increase from 2004 to 2014, with an overall growth tendency (Dai, Huang, Wu, & Zhao, 2016; Zhan, Deng, Wei, Li, & Chen, 2017). From the perspective of vegetation, the annual cumulative NDVI of permanent grassland in Inner Mongolia has increased to over 77.10% from 2002 to 2014 (Gong et al., 2015). The grassland area was reduced from 1.35 billion mu in the 1950s to 800 million mu at the beginning of the 20th century due to excessive use and other factors, such as a decline in available resources and grassland productivity, desertification, and loss of species diversity (Han, Wang, et al., 2018a).
Since 2000, some cultivated land has been transformed to grassland, with the most significant changes in 2004 (Tong et al., 2018). (b) To analyze the degradation factors of grassland ecosystems. The major drivers of grassland degradation in Inner Mongolia are climate change, overgrazing, inadequate policies, and land-use conversion (Cao, Yeh, Holden, Qin, & Ren, 2013; Jiang, Xingguo, & Wu, 2006). (c) To explore protection measures for grassland ecosystems. Briske et al. (2015) offered an alternative strategy that increases both livestock production efficiency and net pastoral income by marketing high-quality animal products to an increasing affluent Chinese economy while simultaneously reducing livestock impacts on grasslands. In addition, grazing exclusion, the “Grain to Green Program” (GGP), and the “Three-North Shelterbelt Project” can also effectively curb grassland ecosystem degradation (Dang et al., 2018; Wang, Zhang, Hasi, & Dong, 2010b). Although current studies of grassland ecosystem degradation in Inner Mongolia are detailed at present, we still lack information on the restoration status of grassland ecosystems in this region.
In view of this, the local net production scaling (LNS) method was used to assess grassland degradation and to reveal the degree to which such ecosystems can be restored. This method is used to determine land degradation and has been proposed by (Prince, 2004) to evaluate land desertification in southern Africa. The method assumes that each area has a potential optimal productivity; due to the joint influences of natural and human factors, there is a gap between the actual and the potential productivity, namely land degradation (Prince, Becker-Reshef, & Rishmawi, 2009). This method can objectively, and accurately, evaluate land degradation and restoration, thereby providing a good explanation for regional land degradation (Noojipady, Prince, & Rishmawi, 2016; Wessels, Prince, & Reshef, 2008). However, this approach has mostly been used in arid areas rather than in more complex climates and terrains. Consequently, the potential productivity in non-arid areas may be misestimated.
In this context, our objectives were as follows: (a) To improve the LNS method so that it can be used in the assessment of grassland ecosystem degradation; (b) to identify the degradation/restoration degree of grassland ecosystems using the advanced local net production scaling method (ALNS). We selected Inner Mongolia to assess grassland degradation, using the methodological theory of the LSN method. Subsequently, we modified the statistical method and constructed the ALNS method to assess grassland degradation in Inner Mongolia. Finally, we revealed and analyzed the spatial degradation degree of grassland in Inner Mongolia to understand the potential changes.
2 STUDY AREA
Inner Mongolia is located in northern China (Figure 1), east of the Great Hinggan Mountains and west of the Ulanqab Plateau, across the semihumid region, the semiarid region, and the arid region in the temperate zone. This area is located between 37°24′–53°23′N and 97°12′–126°04′E, covering 1.146 million km2, at an elevation of 89–3,325 m above sea level, gradually increasing in altitude from east to west. The climate is a typical continental monsoon climate, with cold and long winters and hot and dry summers. Annual precipitation ranges between 50 and 450 mm, gradually increasing from west to east, and mainly occurs in summer. Annual temperature ranges between −3.9 and 8.9°C. From northeast to southwest, the soil types change from Chernozem, chestnut, and meadow to brown calcareous soil (Hoffmann, Funk, Wieland, Li, & Sommer, 2008). The main vegetation type is grassland, from northeast to southwest, the grasslands are meadow grassland, typical grassland, and desert grassland, playing a crucial role in China's ecological sustainability and social and economic development (Dong et al., 2014; Q. Fang, Wang, Liu, Xue, & Yinglan, 2018).
3 RESEARCH METHODS AND DATA SOURCES
3.1 Research methods
3.1.1 ALNS method
Grassland degradation is an aspect of land degradation, they are both influenced by climate change and human activities. Without the impacts of climatic and anthropogenic factors, grassland ecosystem quality solely depends on the natural and geographical conditions (topography, soil, climate, etc.). In view of this, we used the idea of the LNS method and tested the following two hypotheses: (a) In an ecosystem with an ideal state, productivity, function, biomass, and other factors are optimal. Due to the impacts of climate change and human activities, there is a gap between the actual state and the ideal state in terms of the structure and function of the ecosystem, and we defined this gap as ecosystem degradation. (b) Regions within the same ecosystem are heterogeneous, and due to different natural conditions, such as soil, climate, and topography, there are different ideal states within the same ecosystem; on the contrary, regions with the same natural conditions in this ecosystem are homogeneous; hence an ecosystem has different ideal states along with different degrees of degradation.
Based on the above two hypotheses, the definition of degradation, which is different from the previous inherent definition, is the degree to which the actual state deviates from the ideal state, namely the deviation degree, thereby also indicating the degree to which the ecosystem can be recovered.
We performed the following specific research steps: First, we identified and classified the homogeneous regions in the grassland ecosystem; second, we evaluated the ideal ecosystem of each homogeneous region, followed by a quantitative assessment of the degradation degree.
3.1.2 Classification of homogeneous regions
We defined the homogeneous region as a region with similar soil, topography, and climatic conditions. Thus, the related four factors, soil type, precipitation, temperature, and slope, were selected as classification factors to classify the homogeneous regions (Figure 2), using self-organizing feature maps (SOFM) neural network model. This method consists of a competitive learning clustering network, proposed by the Finnish scientist, Kohonen, in 1981, according to the self-organizing characteristics of the human brain, and represents an unsupervised classification method (Kohonen, 1982). The network model conducts competitive learning with input samples. As the inputs with the same functions are close to each other, and the inputs with different functions are separated from each other, some irregular inputs can be automatically ranked (Kohonen, 1990; Vesanto & Alhoniemi, 2000), and the SOFM method can cluster objects adaptively, and self-organized, to ensure the objectivity of the classification results, which is suitable for solving various classification and identification problems (Shon & Moon, 2007; Stankiewicz & Kosiba, 2009).
The purpose of homogeneous region division is to highlight the differences between homogeneous regions and the similarities within each homogeneous region. Too many homogeneous regions will weaken the differences between different regions, while too few homogeneous regions will weaken the similarities within each homogenous region. Thus, an appropriate category number of a homogeneous region is the key to homogeneous region classification. To judge the optimal category number, the clustering quality index (CQI), proposed by Qi et al. (2019), is applied to filter the optimal category numbers. The smaller the CQI, the more appropriate the category number.
((1))
((2))3.1.3 Evaluation of the ideal ecosystem
As the basis of material circulation and energy flow in grassland ecosystems, NPP is the most direct indicator (Dangal, Tian, Chaoqun, Pan, & Pederson, 2016) and was, therefore, selected to characterize and quantify the ideal state of grassland ecosystem. A high NPP value indicates a high ecosystem productivity; therefore, the maximum NPP value represents an ideal grassland ecosystem.
The frequency analysis method is applied to filter the maximum NPP of the ideal state of grassland ecosystem, the so-called 'ideal NPP' (IN) for each homogeneous region. To ensure the accuracy of the IN, the annual mean NPP spatial data from 2000 to 2014 were used to calculate IN. To eliminate extreme outliers of NPP, an NPP value at a specified percentile of frequency was used to represent IN. Typically, the frequency distribution of the actual NPP in each homogeneous region is complex; the NPP at the specified percentile of frequency of each homogenous region is either too large or too small, resulting in the inability to effectively eliminate extreme outliers; alternatively, the INs are underestimated, making the INs deviate from the actual condition.
In this sense, to ensure the rationality of INs, the following two conditions should be satisfied: (a) The number of samples in the specified percentile of frequency where the IN is, must equal or exceed the 1/1,000 of the total samples for each homogeneous region; (b) in all percentiles of frequency that satisfy condition 1, the largest one is the optimal percentile of frequency where the NPP is, representing the IN, and to guarantee the IN is derived from the actual NPP, the IN is the average NPP of all samples in the optimal percentile of frequency.
3.1.4 Assessment of grassland ecosystem degradation
((3))The larger the DN, the greater the degradation degree; when the DN is negative, the grassland ecosystem is the ideal state.
3.2 Data sources
- Soil type data. Soil type data were obtained from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (http://www.resdc.cn/), generated digitally according to the 1:1 Million Soil Map of the People's Republic of China, compiled and published by the National Soil Survey Office of China in 1995 and with a spatial resolution of 1,000 m. The data are classified according to the Chinese Soil Classification System, belonging to the genetic classification system and dividing Chinese soil into 12 orders, 61 groups, and 227 subgroups. The soil data were divided into 34 groups according to the Chinese Soil Classification System.
- Meteorological data. These data, including the annual average precipitation and temperature data from 2000 to 2014 of 39 climate stations (the distribution of climate stations is shown in Figure 1), were downloaded from the China Meteorological Data (http://data.cma.cn/) and then interpolated into a 1,000 × 1,000-m grid via the Kriging interpolation method.
- Slope data. Slope data were generated from the Digital elevation model (DEM). The DEM data were downloaded from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (http://www.resdc.cn/), with a spatial resolution of 1,000 m.
- NPP data. These data were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) Net Primary Productivity (NPP) product (MOD17A3) for 2000–2014, downloaded from NASA (http://ladsweb.nascom.nasa.gov/data/search.html) and acquired on the basis of the MODIS sensor on the TERRA satellite, calculated by the MODIS terrestrial research group with the Biome-BGC (biome-biogeochemical cycles) model with a spatial resolution of 500 m (Foley, 1994). In this study, the NPP data were resampled to 1,000 m. The accuracy of MODIS NPP data has been comprehensively validated in previous studies (Hasenauer, Petritsch, Zhao, Boisvenue, & Running, 2012; Turner et al., 2006). In particular, Peng et al. (2016) used 51 field-measured data to validate MODIS NPP in the Three-North Shelterbelt Program (TNSP) zone (covering Inner Mongolia) and found that it was generally consistent with field-measured NPP, with a correlation coefficient (R) of 0.66 (p < .001) and an RMSE value of 172.9 g C/m2/yr for the overall dataset.
- Land-use data. Land-use data were obtained from the Remote Sensing Monitoring Database of Land Use Status, managed by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (http://www.resdc.cn/). These data use the LANDSAT TM/ETM images as the main data source and are generated by artificial visual interpretation, which started in 1990 with a temporal resolution of 5 years and a spatial resolution of 1,000 m. To avoid the effects of land use change on grassland ecosystem degradation, the no-change grassland areas from 2000 to 2015 were treated as the extents of the grassland ecosystem in Inner Mongolia.
4 RESULTS AND ANALYSIS
4.1 Statistical analysis
Based on the above methods, MATLAB and ArcGIS were applied for data processing and statistical analysis, with the following steps:
First, we interpolated/resampled/calculated the data to 1,000 × 1,000-m in ArcGIS, followed by conversion to the TIFF format for backup.
Second, grassland ecosystem data and classification factor data were imported into MATLAB, and the SOFM method was used to classify the homogenous regions. To obtain the optimal category number of homogeneous regions, the initial category number was set to three classes and gradually increased to 30 classes, with 28 candidate options; the CQIs of each candidate were calculated (Figure 3). Subsequently, a spatial distribution map of homogenous regions was generated according to homogenous regions classification results (Figure 4).
Third, based on classification results and the AN values (Figure 5a), according to the IN statistical principle (3.1.1), we calculated the IN values of each homogenous region (Figure 6) via the frequency analysis method in MATLAB, and generated the IN spatial distribution map (Figure 5b).
In the final step, we calculated the DN value using formula (3) (Figure 5c), and applied correlation analysis to compare the DN value with the actual changes of NPP (the NPP in 2014 minus the NPP in 2000) to verify the rationality of ND value (5.1.3), using MATLAB.
4.2 Classification of the homogeneous regions
As shown in Figure 3, as the category number increases, the CQI decreases first and then increases broadly. When the category was 17, the CQI was the smallest, which is the optimal classification scheme.
We observed a significant spatial heterogeneity among the HRs in the grassland ecosystem of Inner Mongolia (Figure 4). The differentiation into southwest–northeast of the HRs was dominated by temperature and precipitation, and the HRs were roughly distributed in three strip arcs along the northeast to the southwest gradient. The differentiation in southwest–northeast was mainly dominated by soil type and slope, and the HRs were distributed in irregular and staggered grids along the southwest–northeast gradient. The HRs were gradually fragmented and mixed from northwest to southeast.
4.3 Evaluation of the ideal state of grassland ecosystems
The frequency analysis method was used to assess the IN based on AN values. The spatial distributions of AN differed significantly and varied between 0 and 658.6 gc/(m2·yr), with a gradual increase from northeast to southwest (Figure 5a). Regions with high AN values were mainly distributed in the northeast (HR-1, HR-2, and HR-3), where the AN ranged from 200 to 300 gc/(m2·yr), with an average value of each pixel above 240 gc/(m2·yr) (Table 1). The AN values in the middle (HR-4, HR-5, HR-6, HR-13, HR-14, HR-15, HR-16, and HR-17) were smaller than those in the northeast, although the AN of HR-4 ranged from 150 to 250 gc/(m2·yr); the average AN per pixel ranged between 130 and 210 gc/(m2·yr) (Table 1). Regions with low AN values were mainly distributed in the southwest (HR-7, HR-8, HR-9, HR-10, HR-11, and HR-12), with values ranging from 0 to 150 gc/(m2·yr).
| Homogenous region | AN (gc/(m2·yr)) | IN (gc/(m2·yr)) | DN (gc/(m2·yr)) | Homogenous region | AN (gc/(m2·yr)) | IN (gc/(m2·yr)) | DN (gc/(m2·yr)) |
|---|---|---|---|---|---|---|---|
| 1 | 282.01 | 483.49 | 201.48 | 10 | 28.73 | 150.90 | 122.18 |
| 2 | 276.25 | 467.24 | 190.99 | 11 | 105.22 | 125.62 | 20.40 |
| 3 | 240.00 | 397.48 | 157.48 | 12 | 78.80 | 192.33 | 113.53 |
| 4 | 207.90 | 342.94 | 135.04 | 13 | 138.40 | 219.40 | 81.00 |
| 5 | 183.95 | 309.75 | 125.80 | 14 | 171.83 | 339.63 | 167.80 |
| 6 | 184.74 | 300.97 | 116.22 | 15 | 219.25 | 343.94 | 124.69 |
| 7 | 116.44 | 186.46 | 70.02 | 16 | 166.76 | 264.90 | 98.14 |
| 8 | 80.58 | 196.61 | 116.02 | 17 | 187.17 | 286.73 | 99.56 |
| 9 | 40.05 | 104.81 | 64.77 |
The specified percentiles of frequency differed significantly (Figure 6). The regions in the northeast varied from 70 to 80%. The middle part changed considerably, with values for HR-5, HR-6, and HR-17 above 80%, while those of the other regions ranged between 70 and 80%. The southwest values were low, mostly below 70%, with HR-7 and HR-8 reaching values below 50%.
The spatial distribution of INs was similar to that of AN, with values of 100–500 gc/(m2·yr), gradually increasing from southwest to northeast (Figure 5b). Regions with low INs were mainly distributed in the southwest, where the INs ranged from 100 to 200 gc/(m2·yr). In the middle area, the INs of the regions were mainly between 200 and 350 gc/(m2·yr). Regions with high INs were mainly distributed in the northeast, with values exceeding 350 gc/(m2·yr) (Table 1).
4.4 Degradation of grassland ecosystems
The assessment results showed that most grassland areas of Inner Mongolia did not reach their ideal state (about 98.5% of total grasslands), with various degrees of degradation. Only a few areas in the centre of the study area reached the ideal state (1.5% of total grasslands) (Figure5c).
The areas of ideal state were mainly distributed in the regions HR-4, HR-9, and HR-11 of the middle part and in the fragmented parts of HR-1, HR-2, and HR-3 of the northeast. The most degraded areas were concentrated in HR-1, HR-2, and HR-3 of the northeast and in the fragmented parts of HR-4, HR-5, and HR-6 of the midland, with DNs exceeding 200 gc/(m2·yr). In addition, the DNs of most parts of HR-14 and HR-15 in the middle and of HR-8, HR-12, and HR-12 in the southwest ranged from 150 to 200 gc/(m2·yr).
The DNs of other areas were below 150 gc/(m2·yr), and the DNs of HR-13, HR-16, and HR-17 varied from 50 to 100 gc/(m2·yr), with an average of less than 100 gc/(m2·yr). The slightly degraded areas were concentrated in regions HR-7, HR-9, and HR-11 in the southwest, with DN values ranging from 0 from 100 gc/(m2·yr); the average DNs per pixel of these regions were less than 80 gc/(m2·yr).
Overall, grassland degradation was most significant in the northeast, with DN values exceeding 200 gc/(m2·yr), followed by the midland, with values ranging from 50 to 200 gc/(m2·yr). In contrast, grassland degradation in the southwest was negligible, with DN values between 0 and 150 gc/(m2·yr).
5 DISCUSSION
5.1 Analysis of the ALNS method
In this paper, an advanced LNS method was constructed to assess the degradation of grassland ecosystems, and the improvements include the method of classification and frequency analysis.
5.1.1 Comparison of classification methods
Regarding the classification methods used for LNS, researchers have made different attempts. In the original edition of the LNS, designed by (Prince, 2004), the superposition and intersection method (SIM) was used to classify homogenous regions by superposing and intersecting the land and precipitation classes. Noojipady et al. (2016) and Wessels et al. (2008) subsequently used the Analytic Hierarchy Process (AHP) method, K-prototypes, and Iterative Self-Organizing Data Analysis Technique (ISODATA) to classify homogenous regions. As shown in Table 2, each of the four methods has its own advantages and disadvantages. In brief, the SIM and AHP methods are more subjective than the K-prototypes and the ISODATA method, while the latter two are more suitable for classification (Abu Abbas, 2008; Sehgal & Garg, 2014). The K-prototypes and the ISODATA method represent the partitional clustering method, which is relatively sensitive to outliers (Ahmad, 2014).
| Method | Advantages | Disadvantages |
|---|---|---|
| Superposition and intersection method (Prince, 2004) |
|
|
| AHP method (Noojipady et al., 2016; Xu & Xu, 2020) |
|
|
| K-prototype clustering (Akay & Yuksel, 2018; Ji, Bai, Zhou, Ma, & Wang, 2013; Ji, Pang, Zhou, Han, & Wang, 2012; Prince et al., 2009) |
|
|
| ISODATA clustering (Dhodhi, Saghri, Ahmad, & Ul-Mustafa, 1999; Memarsadeghi, Mount, Netanyahu, & Moigne, 2007) |
|
|
In view of this, we attempted to use the SOFM method for classification. This method is superior to the partition clustering method; in that, it retains the original topological structure of the classification factors and is useful for visualizing low-dimensional views of high-dimensional data (De, Chakraborty, & Chakrabarti, 2012; Pratiwi, 2012). Compared with other quantitative clustering methods, this method has shown an advantage in dealing with complex system problems and in describing natural phenomena and laws (Desa, Daeid, Ismail, & Savage, 2010; Kumar, 2015; Rodriguez et al., 2016). This provides a more robust method for classification, which has been applied widely in geography and ecology.
The division of homogenous regions almost coincided with that of the grassland types, illustrating the internal differences of the same grassland types. In addition, the classification result was highly similar to that of the Grassland Classification (NY/T 2997-2016) (2016) issued by the Ministry of Agriculture and Rural Affairs of the People's Republic of China. This not only proved the feasibility and effectiveness of the SOFM method, but also indicated the suitability of the four classification factors. Overall, the SOFM method can provide an objective classification and is generally superior to the original classification method.
5.1.2 Improvement of IN evaluation methods
The LNS method uses the 90th percentile of the AN as the IN in the same homogenous region (Prince, 2004), and other studies have followed this approach. However in this study, this method circumvents the problems caused by the original calculation method; in that, the range of AN of each homogenous region is large and the frequency histogram is not normally distributed. The AN has extreme outliers, which will lead an inaccurate IN by using the specific percentile, and, therefore, the applicability of this method needs to be strengthened. In ideal ecosystem assessment, we used an advanced method to calculate the IN value of each homogenous region, which can effectively eliminate extreme outliers.
By observing the frequency histogram of each homogenous region, we found that the grid numbers of 17 homogenous regions were between 10,000 and 60,000, while those of extreme outliers were between 10 and 60. We, therefore, selected the threshold of 1/1,000 to eliminate the extreme outliers and to calculate a reasonable IN for each homogenous region. We believe that the assessment result has a good accuracy and explanatory power.
5.1.3 Accuracy verification
In this study, degradation is defined as the degree to which the actual state deviates from the ideal state, assessing the degradation of grassland ecosystems by comparing AN and IN values. According to the assessment results, most grassland areas of Inner Mongolia have not reached the ideal state, with varying degrees of variation. We used the actual changes in NPP (ACN) to verify the results. Spatially, the ACN values of most areas increased from 2000 to 2014, and, compared with the degradation results, the regions with a higher ACN growth have a lower degradation degree. Based on the analysis of 2,000 randomly selected samples, ACN growth was positively correlated with the DN, with a correlation coefficient (R) of 0.6716 (Figure 7), and p = .000 < .01 was statistically significant (p = .000). Therefore, the DN values were consistent with the actual changes of the grassland ecosystem, which illustrated that the results of this study can well explain the degradation of grassland ecosystems in Inner Mongolia.
Overall, the advancement of the LNS enhanced the validity and broadened the application scope of the method (the original LNS method was used in arid regions, and, after improvement, it can be used to study ecosystem degradation in arid and semi-arid regions). The ALNS method is suitable to assess ecosystem degradation and can supplement other methods, and our results provide a scientific basis for the development of ecological protection and restoration measures.
5.2 Policy suggestions
Due to the trade-off relationship between economic development and ecological protection, destruction and protection of grassland ecosystems occur simultaneously, which is the reason why grassland ecosystem degradation is still serious despite the implementation of a large number of adequate policies (Dang et al., 2018; Li, Zhang, Shen, Kong, & Zhou, 2020). Therefore, regarding the restoration and protection of grassland ecosystems, we need to pay attention to both development and protection (Lang & Song, 2018, 2019). Our results show that large grassland areas in Inner Mongolia are degraded, which means that they are not in an ideal state and have a high recovery potential, but the degree of recovery does not represent the difficulty degree of recovery. In addition, similar to the ideal ecosystem state, the degree of recovery is also an ideal state, but describes a reference value for ecosystem protection and restoration rather than an actual value that must be reached. Therefore, the DN values could serve as reference values for ecological projects to enhance their scientificity, applicability, and effectiveness.
Specifically, based on the AN and IN values of different grassland types, the ecosystem with the highest quality is meadow grassland, followed by typical grassland, while desert grassland has the lowest values. The higher the degradation degree of the high-quality grassland type, the greater its recoverability and its actual growth (Table 3). It is worth noting that, based on the degradation rate (the ratio of DN to IN) of different grassland types, the degradation rate of desert grassland reached 0.6, which deviated from the ideal state to a greater extent, with more serious degradation. Grassland degradation is mainly characterized by the degradation of grassland coverage, and desertification has occurred in Inner Mongolia (Li, Yan, Wang, & Du, 2019; Zhang, Wang, Wang, Yang, & Li, 2020). Due to the differences in soil condition and vegetation composition, desert grassland is sensitive to climate change and human activities and, therefore, more vulnerable (Cai, Yang, & Xu, 2015). Ecological projects have different effects on different ecological systems (Cao, Ma, Yuan, & Wang, 2014; Jiyuan, Xinliang, & Quanqin, 2008), and, in this sense, ecological protection and restoration projects should be more detailed and targeted.
| Grassland type | IN (gc/(m2·yr)) | AN (gc/(m2·yr)) | DN (gc/(m2·yr)) | Degradation rate |
|---|---|---|---|---|
| Meadow grassland | 360.98 | 225.63 | 135.35 | 0.38 |
| Typical grassland | 275.52 | 162.74 | 112.78 | 0.41 |
| Desert grassland | 163.96 | 65.49 | 98.47 | 0.60 |
- Note: The degradation rate is the ratio of DN to IN.
6 CONCLUSIONS
We evaluated the degradation of grassland ecosystems in Inner Mongolia, using the ALNS method. Based on the results, we can draw the following conclusions: (a) The ALNS method divides the grassland system into 17 homogeneous regions, with the INs ranging from 100 to 500 gc/(m2·yr), increasing from the southwest to the northeast, suggesting that the quality of grassland ecosystems is higher in the northeast. (b) Most regions of the grassland ecosystem failed to reach an ideal state and were degraded to varying degrees, with the northeast being the most significant area, followed by the midland; degradation in the southwest was insignificant. The LANS method could provide a new approach to estimate grassland ecosystem degradation and restoration.
ACKNOWLEDGMENTS
The research was supported by the National Key Research and Development Program of China (Grant No. 2017YFC0503803) and the National Natural Science Foundation of China (Grant No. 41771460).
CONFLICT OF INTEREST
The authors declare no conflicts of interest.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
