Response of Reduced Grassland Degradation Index to Climate Change in China

Abstract

Grasslands have been increasingly impacted by human activities, gradually becoming one of the most threatened ecosystems globally. Advanced geographic information technology and remote sensing techniques allow for a fresh perspective on studying the response of the grassland degradation index ($GDI$) to climate change. This study utilized remote sensing image data of grasslands to calculate the vegetation coverage and derive the $GDI$ for five grassland regions of China from 2001 to 2019. The results indicate that the national degradation status of grasslands remained at a level of mild degradation. The increasing trend of the $GDI$ in some regions was effectively inhibited by regional climate change, especially in the Northeastern and Northern Plain–Mountain–Hill Grassland regions, where the $GDI$ showed a continuous decreasing trend. $GDI$ was strongly correlated with atmospheric pressure, precipitation, temperature, and wind speed. In the arid northern region, the increasing precipitation and decreasing temperatures predominantly contributed to the depressed $GDI$. In the Qinghai–Tibetan Plateau Grassland region, the instability of the $GDI$ is attributed to fluctuating atmospheric pressure, with a correlation coefficient ranging from 0.5 to 0.8. Our findings underscore the importance of meteorological factors to evaluate and forecast grassland ecosystem stability. This understanding is vital for developing informed conservation and management strategies to address current and future climate challenges.

1. Introduction

Grasslands, the second-largest terrestrial ecosystem, occupy over 40% of the global land area and play a crucial role in livestock production and environmental conservation [1]. They offer numerous ecological services such as supporting biodiversity, regulating climate, and preventing soil erosion [2]. The global temperature has been on a continuous increasing trend since the 1970s. The Intergovernmental Panel on Climate Change’s (IPCC) Sixth Assessment Report (AR6) from Working Group II highlights a significant increase in global river and lake temperatures since the 1970s, with increments of 1 °C and 0.45 °C per decade, respectively [3]. Global environmental changes, marked by climate warming, have triggered a series of significant issues worldwide. These include rising sea levels, land degradation, desertification, ecosystem degradation, drastic reductions in biodiversity, and extreme climate events [4]. These changes will lead to dramatic alterations in the Earth’s environment, upon which human survival and development depend. Climate warming has significant impacts on terrestrial ecosystems, with grasslands being a crucial and particularly vulnerable component. The degradation of grasslands significantly impacts grassland productivity and ecosystems, drawing considerable attention to the challenges faced by fragile and unstable grassland farming [5]. Grasslands have become one of the most endangered ecosystems worldwide, increasingly threatened by various human activities. As a result, many countries and regions face serious challenges of grassland degradation [2,5,6]. According to a survey, the global grassland area is approximately 3.2 billion hectares, with approximately 49% being degraded [7,8]. Grassland degradation refers to the process where natural grasslands deteriorate due to adverse natural factors like drought, wind erosion, water erosion, salinization, waterlogging, and changes in groundwater levels [3]; irrational utilization such as overgrazing and excessive mowing [4]; or human activities like over-exploitation and deforestation that destroy vegetation cover [5]. The study shows that from 2000 to 2022, Africa and Asia ranked first and second in global grassland degradation [9], with degradation levels of 19.58% and 39.35%, respectively [1]. Central Asia experienced a degradation rate of 23.08% between 2000 and 2020, with 53.8% attributed to climate change and 14.5% to human activities [10,11,12]. In China, the grasslands of the Qinghai–Tibet Plateau have significantly degraded [13], with a proportion of 12.63% from 2000 to 2020 [14,15]. In Inner Mongolia, the Hulunbuir grasslands, for instance, exhibited a degradation level above moderate ($GDI$ > 2) between 2003 and 2012, fluctuating between moderate and severe degradation with dramatic interannual changes in $GDI$ [16,17,18]. From 1986 to 2000, the grassland area in the western Songnen Plain decreased by 24.7%, with severe degradation transforming grasslands mainly into forests, farmlands, idle lands, and saline–alkali lands [19]. As grassland quality declines, the numbers of species and livestock have increased significantly [20].

Climate change is one of the potential driving factors of grassland degradation [21], while high levels of biodiversity can endow grassland ecosystems with stability, reducing their susceptibility to environmental changes. Additionally, biodiversity plays a crucial role in influencing the quality of grassland productivity [22,23]. Based on comparative experiments of grassland ecosystems, biodiversity plays a crucial role in ecosystem functionality, indicating that productivity increases with increasing biodiversity [24,25]. Research has shown that drought can have a significant impact on grassland growth [26]. Drought can lead to insufficient soil moisture for plant growth, with high rates of soil moisture loss due to evapotranspiration. This can result in water stress for plants early in the growing season, leading to reduced plant yield and significant fluctuations in grassland productivity [27,28]. Zhang et al. suggested a close relationship between the increase in soil temperature and grassland degradation. Higher soil temperatures indicate the weaker ability of grasslands to intercept solar radiation, leading to higher levels of degradation. On the other hand, higher soil temperatures also create more favorable conditions for the survival of insects and disease during winter, which exacerbates grassland degradation [29]. Therefore, regional-scale monitoring and assessment of grassland degradation, as well as grassland restoration and reconstruction, have become important aspects of scientific research on global change and sustainability. Tueller et al. utilized AVHRR data to estimate the grassland resources covering an area of 5.2 × 10⁴ km² in the desertification area of Nebraska, USA, providing a scientific basis for the rational utilization of grasslands as early as 1989 [30,31,32]. Ringrose [33,34] and A. Karnieli et al. [35,36,37], respectively, used remote sensing methods to study the degradation of southeastern grasslands in Botswana and land degradation in arid regions of Kazakhstan during 1978–1987. Fan Ying characterized vegetation cover in the western Inner Mongolia Plateau from 2000 to 2012 by analyzing MODIS EVI data and investigated the improvement and degradation status of desert areas, grasslands, meadows, and shrublands [38,39]. Qian et al. utilized a natural grassland classification method based on decision tree models, using the normalized difference vegetation index ($NDVI$) and digital elevation model (DEM) obtained from TM remote sensing images as parameters for the decision tree model. They classified the desert grasslands on the northern slope of the Tianshan Mountains and achieved good accuracy [40,41,42].

Regarding research on the grassland degradation index, studies to date have been confined to regional areas and qualitative analysis of the relationship between $GDI$ and climate change. Clearly, an improved understanding of the $GDI$ changes at a larger scale and its response to climate variation through quantitative analysis can greatly enhance our knowledge of the relationships between ecosystem vegetation and climate change.

This study utilizes data, incorporating the normalized difference vegetation index ($NDVI$) and observational climate data from 473 stations across China, to explore spatiotemporal variations of $NDVI$ and their responses to climatic variations in grassland categorized into five ecological functional regions from 1998 to 2020. The objective is to enhance our understanding and predictive capabilities regarding grassland degeneration across China. By illuminating the intricate relationship between vegetation and climate change in this globally widespread area, the findings of our study can offer valuable new insights for ecological management and conservation efforts.

2. Materials and Methods

2.1. Study Area

China is located in eastern Asia, bordering the western Pacific Ocean (73°33′ E to 135°05′ E, 3°51′ N to 53°33′ N), and is characterized by a complex and diverse range of climatic types. In order to investigate the spatiotemporal change of grassland degradation in different ecological functional conditions, grasslands in China were divided into five regions according to ecological function. These regions include the Inner Mongolia Plateau grassland region (IM), Northwestern Mountain–Basin grassland region (NM), Qinghai–Tibetan Plateau grassland region (TP), Northeastern and Northern Plain–Mountain–Hill grassland region (NN), and Southern Mountain–Hill grassland region (SM) (Figure 1).

2.2. Data Sources

2.2.1. Remote Sensing Data

Satellite-derived $NDVI$ data covering the period from 1998 to 2019 were obtained from the MOD09A1 $NDVI$ dataset, with temporal and spatial resolutions of 8 days and 250 m, respectively [31]. This dataset was provided by the Earth Science Data Systems of the National Aeronautics and Space Administration. The distribution of grassland in the study area was obtained from the vegetation regionalization datasets at a scale of 1:1,000,000 for China, sourced from the Resource and Environmental Science Data Center for registration and publication. The study used the 8-day composite product of MODIS 250 m resolution reflectance (MOD09A1), encompassing seven different spectral bands. To ensure the highest data quality, a maximum value composite method was adopted, analyzing the pixel with the highest reflectance value within every 8-day period. From 2001 to 2019, a total of 35 images were selected, spanning from March to November each year. Reflectance data for the red and near-infrared bands were extracted using the MODIS Conversion Toolkit (MCTK) module in ENVI 5.1 software. The images were then reprojected using Delaunay triangulation for nearest neighbor sampling to convert them into UTM projection with a 51N region. From these images, the Normalized Difference Vegetation Index ($NDVI$) was calculated. $NDVI$ images were then cropped according to geographical coordinates to obtain a long-term $NDVI$ dataset covering the entirety of China.

2.2.2. Meteorological Data

The meteorological data used in this research were extracted from the China Surface Climate Data Daily Data set (Version 3.0). These observations were obtained from the China Meteorological Data Sharing Service System (https://data.cma.cn//) accessed on 1 May 2023. A total of 231 meteorological stations were distributed across five grasslands regions.

These included 11 climate factor daily records, mean air temperature ($A{T}_{mean}$), maximum air temperature ($A{T}_{\max }$), minimum air temperature ($A{T}_{\min }$), mean surface temperature ($S{T}_{mean}$), maximum surface temperature ($S{T}_{\max }$), minimum surface temperature ($S{T}_{\min }$), 24 h precipitation ($PR$), mean station atmospheric pressure ($AP$), average relative humidity ($HU$), mean wind speed ($WS$), and sunshine duration ($SD$). This was designed as such to maintain consistency with the study period of the $NDVI$ data from 2001 to 2019.

2.3. Data Processing and Analysis

2.3.1. $GDI$ Change

Using vegetation cover to indicate the extent of grassland degradation has emerged as a prevalent detection technique in the field of remote sensing in recent years. This study generated a Normalized Difference Vegetation Index ($NDVI$) dataset for the research area using MODIS data with a spatial resolution of 250 m. The data covered the growing seasons from March to November, spanning from 2001 to 2019. Based on this dataset, vegetation cover was calculated for each of the five grassland regions in China. The specific formula used for this calculation is as follows:

$$V_c={\displaystyle \frac{NDVI-NDV{I}_s}{NDV{I}_v-NDV{I}_s}}$$

(1)

In the equation, $V_c$ represents vegetation cover; $NDVI$, $NDV{I}_{\mathrm{s}}$, and $NDV{I}_{\mathrm{v}}$ correspond to the pixel’s $NDVI$ value, the minimum $NDVI$ value for bare soil, and the maximum $NDVI$ value for pure grassland, respectively. The maximum vegetation cover calculated for each pixel serves as the vegetation cover for undegraded grassland during the period of 1998–2000. This paper establishes grading standards based on the findings of recent studies in similar regions, as illustrated in

Table 1. Grassland degradation level classification method.

Degradation Level	Score	Classification Method
Non-degraded	1	Grassland vegetation cover exceeds 90% of V *
Mild-degraded	2	Grassland vegetation cover ranges from 75% to 90% of V *
Moderate-degraded	3	Grassland vegetation cover ranges from 60% to 75% of V *
Severe-degraded	4	Grassland vegetation cover ranges from 30% to 60% of V *
Extreme serious-degraded	5	Grassland vegetation cover is below 30% of V *

* V represents the vegetation cover of non-degraded grassland.

, and utilizes these standards to classify the extent of grassland degradation from 2001 to 2019.

Considering that this classification method does not incorporate the variable of degraded grassland area in reflecting the spatial distribution of grassland degradation, for the sake of accuracy and clarity in the research, this study integrates the results of previous research and uses the grassland degradation index ($GDI$) to characterize the condition of grassland degradation across the entire area. The specific values of the $GDI$ are presented through line graphs generated by SigmaPlot 14.0 software. The formula for calculating the grassland degradation index is as follows:

$$GDI=({\displaystyle \sum _{i=1}^5{D}_i\times A_i)}/A$$

(2)

In the formula, $GDI$ represents the grassland degradation index; $D_i$ is the score for grassland degradation level $\mathrm{i}$ (with the scores for each pixel’s degradation level presented in Table 1); $A_i$ is the area of grassland at degradation level $\mathrm{i}$; and $A$ is the total area of $A_i$. Higher values of $GDI$ indicate a more severe degree of grassland degradation in the area. The $GDI$ (ranging from 1 to 5) indicates the following grassland degradation conditions, respectively (

Table 2. Grassland degradation situation classification method.

GDI	Classification of Grassland
[0, 1)	Non-degraded grassland
[1, 2)	Mild-degraded grassland
[2, 3)	Moderate-degraded grassland
[3, 4)	Severe-degraded grassland
[4, ∞)	Extreme serious-degraded grassland

).

2.3.2. Meteorological Dataset

Daily records of 11 climate factors were processed using SAS 9.4 software to calculate monthly averages, seasonal averages (winter: December of the previous year, January and February of the current year; spring: March, April, and May; summer: June, July, and August; autumn: September, October, and November), and annual average values within the growing season (March to November). This process provided a comprehensive meteorological dataset for the five grassland regions in China for the years 2001–2019.

2.3.3. Correlation Analysis

To investigate the $GDI$ in response to climate variations, we analyzed the simple correlation coefficients between 11 monthly, seasonal, and annual climate factors and the $GDI$. In addition, we carried out a partial correlation analysis (When the correlation between one meteorological factor and $GDI$ is analyzed, the other 10 meteorological factors are taken as control variables.) to further examine correlations between climate variables and the $GDI$. Through this analysis, it was possible to determine the relationship between two parameters after removing the influence of other factors [32,33]. Within SAS 9.4 software, an analysis of correlation between the meteorological data and the grassland degradation index ($GDI$) for the five grassland regions in China from 2001 to 2019 was conducted, yielding a series of correlation and partial correlation values. SigmaPlot 14.0 software was utilized to create bar graphs illustrating the relationship between the annual average data and the $GDI$. Correlation analysis revealed the correlation values between the grassland degradation index over nineteen years and eleven meteorological factors across the five major grassland areas during the four seasons. The data were organized and subsequently visualized through heat maps generated using IBM SPSS Statistics 27 software.

3. Results

3.1. Five Major Grassland Regions in China

The area of grasslands in China is 2.85 × 10⁶ km². Of the five regions, the TP has the largest area of grassland, reaching 15.8 × 10⁵ km², accounting for 55.5% of the total area. The IM and NM regions follow, with areas of 6.35 × 10⁵ km² and 4.47 × 10⁵ km², accounting for 22.3% and 15.7%. The grassland in NN and SM is relatively low, accounting for 5% and 1.5% of the total area.

As shown in Figure 2, the spatial distribution of grassland degradation from 2001 to 2019 exhibited obvious variation. Overall, the grassland degradation in the IM is more pronounced compared to other regions, with most areas experiencing severe degradation. This phenomenon became particularly evident in 2005.

3.2. Spatiotemporal Distribution of Grassland Degradation

The classification results of grassland degradation in the study regions of China from 2001 to 2019 are shown in Figure 2. It can be observed that during this 19-year period, the spatial distribution of grassland degradation in China’s ecological functional regions exhibited obvious variation. Overall, the grassland degradation in the IM is more pronounced compared to other regions, with most areas experiencing severe degradation. In contrast, other regions are primarily characterized by mild to moderate degradation. This phenomenon became particularly evident in 2005.

In the IM, the degree of grassland degradation fluctuated dramatically, making it difficult for the grassland vegetation to achieve sustainable and stable recovery in this ecological region between 2001 and 2019. As depicted in Figure 2, the areas of moderate and severe degradation in IM experienced several sharp increases, followed by gradual declines over periods of 3–4 years between 2001 and 2012. Vegetation requires some recovery time after severe destruction. However, the grasslands continued to remain severely degraded in Xilinhot and the western part of Hailaer, with a spreading trend of severe degradation area from 2014 to 2017. In the NM, the degree of grassland degradation fluctuated between severe, moderate, and mild from 2001 to 2019. However, it essentially reached a non-degraded state in 2013. It is evident from Figure 2 that compared to IM, NM exhibited milder degrees of severe and moderate degradation, with varying degrees of degradation being more dispersed in this region. From 2001 to 2014, there was a noticeable weakening trend in the degree of degradation in NM, despite some spreading of severe and moderate degradation areas during 2008 to 2009. However, this spreading trend was effectively suppressed from 2010 onwards, leading to a continuous reduction in degradation in the area. The TP, being the region with the largest proportion of grassland area in China, maintained a relatively stable non-degraded state from 2001 to 2019. Only small areas of severe and moderate degradation appeared at the border with NM from 2004 to 2006, and these areas reverted to a non-degraded state in 2016. Most of the grasslands in NN are distributed along its boundary with IM, exhibiting a spotted pattern. Due to the adjacency of grasslands in this area to IM, the degree of grassland degradation in this region has also fluctuated unstably over the 19-year period. In contrast, the grasslands in SM are mainly located along its border with TP, also exhibiting a spotted distribution pattern but with a smaller coverage range, appearing to have a more concentrated trend. The grassland degradation in SM has remained at a non-degraded state over the 19-year study period, indicating its strong grassland ecological function. It is capable of maintaining ecological stability under various external influences.

3.3. Temporal Changes of Grassland Degradation Index of China in Past 20 Years

Figure 3 shows the inter-annual variations of grassland degradation index ($GDI$) in five regions. The IM experiences the most drastic fluctuations, generally changing between mild and moderate degradation. The ecological functional performance of this area appears relatively fragile, as evidenced by a sharp increase to the peak value ($GDI$ = 2.12) in just one year (2008–2009) followed by a decrease to the lowest value ($GDI$ = 1.13) over the subsequent three years. The change of $GDI$ in NM maintained a pattern similar to that in IM during the study period. However, it revealed relatively gentle fluctuations of values, verifying that Ningxia remained consistently at a mild degradation level over the 20-year period. It remains between mild-degraded states and moderate-degraded states. $GDI$ in the TP maintained a steady declining trend from 2001 to 2013 at the rate of 0.13 per decade, then entered into a relatively irregular upward phase after reaching its lowest value ($GDI$ = 1.06) in 2013. It reached its highest value of 1.42 in 2014–2015. Compared with other regions, the inter-annual variation of NN was relatively flat and showed a continuous downward trend (a = −0.13). The variation of NN has a high matching degree with the trend line (R = 0.75), showing an obvious degree of fit from 2001 to 2019. Although $GDI$ showed a slight increase in 2005–2006 and 2014–2015, the survey data in 2019 ($GDI$ = 1.03) was obviously lower than that at the beginning of the study ($GDI$ = 1.52). This suggests that the grassland in this area has been in a relatively stable state of continuous recovery in the past 19 years. There were slight variations (between 1.01 and 1.06) in mildly degraded states throughout 20 years in SM, indicating stable grassland ecological functional performance.

At the national level, the overall grassland degradation situation in China has remained at the level of mild degradation ($GDI$ = 1.31) over the past 20 years. The comparison between each subgraph in Figure 3 revealed that although the trends of $GDI$ variations differ among the five grassland regions, all the grassland showed an upward trend in 2013–2014, with IM and NM experiencing the most obvious increases. Regions experiencing rapid increases in $GDI$ over a short period followed by rapid decreases in the following 2–4 years indicate no regions with steady increases in $GDI$ over the whole study period. From a spatial perspective, the northern regions, especially IM, exhibited severe degradation with extensive areas affected. In contrast, the western regions experienced less severe degradation compared to the eastern regions, predominantly characterized by mild to moderate degradation before 2013.

3.4. Response of Grassland Degradation Index to Climate Change

3.4.1. Relationships between the $GDI$ and Annual Climate Variables

We first analyzed simple correlations between the annual mean value of climate variables and the $GDI$ in five regions.

In the IM, the $GDI$ showed a positive simple correlation with $S{T}_{mean}$, $S{T}_{\max }$, and $A{T}_{mean}$, with correlation coefficients of 0.50, 0.46, and 0.46, respectively, demonstrating that the increase in temperature has a certain inhibitory effect on the growth of vegetation. Furthermore, a positive correlation was discovered between $GDI$ and $PR$, with the correlation coefficient being −0.73, indicating a remarkable influence of $AP$ on $GDI$. The changes in temperature can lead to uneven water distribution in grasslands, consequently resulting in grassland degradation.

In the NM, $PR$ had a significant negative simple correlation with $GDI$ (r = −0.68). Meanwhile, the $GDI$ showed a positive significant correlation with $S{T}_{\max }$, with a correlation coefficient of 0.47. However, the partial correlations between $GDI$ and all of the meteorological factors were not statistically significant (p > 0.5). This lack of significance can be attributed to the promotion effect of $PR$ on vegetation cover and the promotion effect of $S{T}_{\max }$ on $GDI$ canceling each other out, resulting in no significant partial correlation between meteorological factors and $GDI$ in this region.

In the TP, the $GDI$ showed a significant positive correlation with $AP$, with a correlation coefficient of 0.63. Since the altitude of TP is mostly above 4000 m, $AP$ in this area is also obviously higher than in other regions, which indicates a weak correlation between other meteorological factors and $GDI$.

In the NN, the $GDI$ showed a significant negative correlation with $PR$ (r = −0.58), $SD$ (r = −0.57), and $S{T}_{\min }$ (r = −0.71). NN has a long winter, and there is a large temperature difference between day and night in this region, which demonstrates that the $GDI$ of NN has the most obvious simple correlation with $S{T}_{\min }$. Additionally, the $GDI$ also showed a significant positive simple correlation with $WS$ (r = 0.60) and $S{T}_{\mathrm{mean}}$ (r = 0.49). This result can be attributed to the fact that $WS$ promotes grassland degradation in most of the plain areas in NN.

In the SM, $WS$ represented a significant negative correlation with $GDI$ (r = −0.52). The $GDI$ showed a significant positive partial correlation with $SD$ (r = 0.68) and showed a significant negative partial correlation with $S{T}_{\max }$ (r = −0.68) and $S{T}_{\min }$ (r = −0.68) (Figure 4).

3.4.2. Relationships between the $GDI$ and Seasonal Climate Variables

In the IM, the $GDI$ showed a significant positive correlation with $S{T}_{\mathrm{mean}}$ and $S{T}_{\max }$ in summer, with correlation coefficients of 0.71 and 0.64, respectively. Notably, in July, they reached 0.73 and 0.71, respectively (

Table A1. Monthly variations in the correlation between the grassland degradation index ($GDI$) and various meteorological factors across the grassland regions in China from 2001 to 2019.

Region	Meteorological Factors	Mar.	Apr.	May	Jun.	Jul.	Aug.	Sep.	Oct.	Nov.
NN	ATmean	−0.08	−0.26	0.67 *	−0.30	−0.16	0.45	0.46	0.09	−0.33
	ATmax	−0.04	0.27	−0.60	0.59	0.18	−0.35	−0.40	−0.34	0.31
	ATmin	0.14	0.27	−0.61	0.23	0.10	−0.36	−0.42	0.38	0.29
	HU	0.33	−0.08	0.53	−0.10	−0.27	0.05	0.38	−0.25	0.53
	PR	−0.35	−0.47	−0.35	−0.48	0.50	−0.18	−0.27	−0.62	−0.19
	WS	0.19	−0.02	0.66	0.40	0.57	−0.17	0.42	0.61	0.26
	AP	−0.28	0.19	−0.22	0.43	0.34	0.37	0.75 *	0.48	0.07
	STmean	0.10	0.04	−0.75 *	−0.65	−0.19	−0.37	−0.36	−0.31	0.31
	SD	0.37	−0.41	0.29	−0.42	0.04	−0.04	−0.47	0.52	0.23
	STmax	−0.05	−0.06	0.69 *	0.61	0.28	0.30	0.30	0.23	0.01
	STmin	0.15	0.24	0.58	0.25	0.12	0.21	0.32	0.07	0.30
SM	ATmean	−0.48	0.31	−0.10	−0.38	0.27	0.21	−0.01	0.11	0.39
	ATmax	0.49	−0.24	−0.08	0.36	−0.13	−0.18	−0.25	−0.09	−0.25
	ATmin	0.46	−0.25	0.18	0.46	−0.45	−0.25	0.22	−0.07	−0.41
	HU	−0.21	0.52	0.15	0.32	0.64	0.01	−0.21	0.27	0.28
	PR	0.24	−0.29	0.08	−0.28	−0.20	0.18	−0.15	−0.19	0.03
	WS	−0.03	0.05	0.22	−0.34	−0.06	0.01	0.08	0.29	−0.10
	AP	0.49	0.20	−0.14	−0.14	0.38	0.12	0.46	0.12	0.14
	STmean	0.48	0.39	0.63	0.37	−0.26	−0.04	0.38	0.18	0.01
	SD	0.39	0.08	0.42	0.28	−0.32	0.08	0.33	0.11	−0.08
	STmax	−0.52	−0.28	−0.54	−0.29	0.47	0.04	−0.37	−0.16	−0.02
	STmin	−0.49	−0.36	−0.72 *	−0.41	0.36	0.07	−0.31	−0.25	−0.02
IM	ATmean	−0.10	0.48	−0.27	0.43	0.31	−0.16	−0.20	−0.38	−0.24
	ATmax	−0.07	−0.48	0.37	−0.47	−0.40	−0.08	0.09	0.04	0.20
	ATmin	0.12	−0.41	0.13	−0.36	−0.11	0.31	0.29	0.57	0.18
	HU	−0.22	0.12	−0.12	0.13	−0.25	−0.42	0.10	−0.53	−0.13
	PR	0.25	0.42	−0.27	0.12	−0.14	0.43	−0.42	0.27	−0.07
	WS	0.25	0.19	0.13	0.05	0.67	0.25	0.06	−0.27	−0.14
	AP	0.14	0.34	0.24	0.17	−0.25	0.46	0.02	−0.08	0.22
	STmean	0.24	−0.25	−0.34	−0.13	−0.19	0.23	0.60	0.60	0.53
	SD	−0.03	−0.16	0.48	−0.25	−0.06	−0.42	−0.19	0.26	0.21
	STmax	−0.15	0.20	0.13	0.17	0.31	0.24	−0.49	−0.54	−0.25
	STmin	−0.26	−0.18	0.23	0.05	−0.09	−0.55	−0.68 *	−0.66	−0.50
TP	ATmean	0.62	−0.45	−0.16	−0.51	0.07	0.32	0.59	0.25	−0.09
	ATmax	−0.62	0.75 *	0.22	0.45	−0.11	−0.18	−0.43	−0.04	0.09
	ATmin	−0.60	−0.22	0.11	0.51	−0.01	−0.41	−0.66	−0.44	0.09
	HU	0.72 *	0.57	−0.08	0.00	−0.22	0.61	0.54	0.56	−0.71 *
	PR	−0.83 *	−0.46	−0.03	0.01	0.28	−0.37	−0.13	0.06	−0.67 *
	WS	0.85 *	0.71 *	0.50	0.38	0.06	0.38	0.60	0.82 *	0.27
	AP	0.80 *	0.02	0.06	0.27	0.14	0.71 *	−0.47	−0.13	0.77 *
	STmean	−0.29	−0.70 *	0.59	0.62	0.34	−0.43	−0.34	−0.13	0.66
	SD	0.18	−0.51	−0.44	0.10	0.22	−0.38	−0.41	−0.31	−0.78 *
	STmax	0.26	0.59	−0.58	−0.59	−0.30	0.46	0.48	0.22	−0.66
	STmin	0.28	0.67 *	−0.58	−0.63	−0.35	0.39	0.31	0.11	−0.63
NM	ATmean	0.12	0.11	0.13	0.15	0.58	−0.39	−0.51	−0.57	0.25
	ATmax	−0.34	0.00	−0.30	0.05	−0.64	0.45	0.42	0.53	−0.14
	ATmin	0.19	−0.29	0.21	−0.17	−0.34	0.31	0.17	0.69 *	−0.31
	HU	−0.50	−0.06	−0.06	−0.45	−0.86 *	0.02	0.80 *	−0.42	0.45
	PR	−0.28	0.25	0.19	−0.36	0.64	0.46	−0.48	0.44	−0.35
	WS	0.76 *	0.17	0.61	−0.52	−0.62	0.22	0.24	0.72 *	−0.10
	AP	−0.10	0.08	−0.14	0.65	0.35	0.43	−0.40	0.05	−0.06
	STmean	−0.13	0.11	0.02	−0.33	−0.48	0.21	0.45	0.26	−0.09
	SD	0.58	−0.42	−0.28	−0.45	−0.81 *	0.05	0.45	0.15	−0.18
	STmax	−0.04	−0.07	0.32	0.37	0.28	−0.16	−0.25	−0.23	0.03
	STmin	0.00	0.00	−0.45	0.29	0.05	−0.30	−0.26	−0.56	0.12
NN	ATmean	−0.08	−0.26	0.67*	−0.30	−0.16	0.45	0.46	0.09	−0.33
	ATmax	−0.04	0.27	−0.60	0.59	0.18	−0.35	−0.40	−0.34	0.31
	ATmin	0.14	0.27	−0.61	0.23	0.10	−0.36	−0.42	0.38	0.29
	HU	0.33	−0.08	0.53	−0.10	−0.27	0.05	0.38	−0.25	0.53
	PR	−0.35	−0.47	−0.35	−0.48	0.50	−0.18	−0.27	−0.62	−0.19
	WS	0.19	−0.02	0.66	0.40	0.57	−0.17	0.42	0.61	0.26
	AP	−0.28	0.19	−0.22	0.43	0.34	0.37	0.75 *	0.48	0.07
	STmean	0.10	0.04	−0.75 *	−0.65	−0.19	−0.37	−0.36	−0.31	0.31
	SD	0.37	−0.41	0.29	−0.42	0.04	−0.04	−0.47	0.52	0.23
	STmax	−0.05	−0.06	0.69 *	0.61	0.28	0.30	0.30	0.23	0.01
	STmin	0.15	0.24	0.58	0.25	0.12	0.21	0.32	0.07	0.30
SM	ATmean	−0.48	0.31	−0.10	−0.38	0.27	0.21	−0.01	0.11	0.39
	ATmax	0.49	−0.24	−0.08	0.36	−0.13	−0.18	−0.25	−0.09	−0.25
	ATmin	0.46	−0.25	0.18	0.46	−0.45	−0.25	0.22	−0.07	−0.41
	HU	−0.21	0.52	0.15	0.32	0.64	0.01	−0.21	0.27	0.28
	PR	0.24	−0.29	0.08	−0.28	−0.20	0.18	−0.15	−0.19	0.03
	WS	−0.03	0.05	0.22	−0.34	−0.06	0.01	0.08	0.29	−0.10
	AP	0.49	0.20	−0.14	−0.14	0.38	0.12	0.46	0.12	0.14
	STmean	0.48	0.39	0.63	0.37	−0.26	−0.04	0.38	0.18	0.01
	SD	0.39	0.08	0.42	0.28	−0.32	0.08	0.33	0.11	−0.08
	STmax	−0.52	−0.28	−0.54	−0.29	0.47	0.04	−0.37	−0.16	−0.02
	STmin	−0.49	−0.36	−0.72 *	−0.41	0.36	0.07	−0.31	−0.25	−0.02
IM	ATmean	−0.10	0.48	−0.27	0.43	0.31	−0.16	−0.20	−0.38	−0.24
	ATmax	−0.07	−0.48	0.37	−0.47	−0.40	−0.08	0.09	0.04	0.20
	ATmin	0.12	−0.41	0.13	−0.36	−0.11	0.31	0.29	0.57	0.18
	HU	−0.22	0.12	−0.12	0.13	−0.25	−0.42	0.10	−0.53	−0.13
	PR	0.25	0.42	−0.27	0.12	−0.14	0.43	−0.42	0.27	−0.07
	WS	0.25	0.19	0.13	0.05	0.67	0.25	0.06	−0.27	−0.14
	AP	0.14	0.34	0.24	0.17	−0.25	0.46	0.02	−0.08	0.22
	STmean	0.24	−0.25	−0.34	−0.13	−0.19	0.23	0.60	0.60	0.53
	SD	−0.03	−0.16	0.48	−0.25	−0.06	−0.42	−0.19	0.26	0.21
	STmax	−0.15	0.20	0.13	0.17	0.31	0.24	−0.49	−0.54	−0.25
	STmin	−0.26	−0.18	0.23	0.05	−0.09	−0.55	−0.68*	−0.66	−0.50
TP	ATmean	0.62	−0.45	−0.16	−0.51	0.07	0.32	0.59	0.25	−0.09
	ATmax	−0.62	0.75 *	0.22	0.45	−0.11	−0.18	−0.43	−0.04	0.09
	ATmin	−0.60	−0.22	0.11	0.51	−0.01	−0.41	−0.66	−0.44	0.09
	HU	0.72 *	0.57	−0.08	0.00	−0.22	0.61	0.54	0.56	−0.71 *
	PR	−0.83 *	−0.46	−0.03	0.01	0.28	−0.37	−0.13	0.06	−0.67 *
	WS	0.85 *	0.71 *	0.50	0.38	0.06	0.38	0.60	0.82 *	0.27
	AP	0.80 *	0.02	0.06	0.27	0.14	0.71 *	−0.47	−0.13	0.77 *
	STmean	−0.29	−0.70 *	0.59	0.62	0.34	−0.43	−0.34	−0.13	0.66
	SD	0.18	−0.51	−0.44	0.10	0.22	−0.38	−0.41	−0.31	−0.78 *
	STmax	0.26	0.59	−0.58	−0.59	−0.30	0.46	0.48	0.22	−0.66
	STmin	0.28	0.67 *	−0.58	−0.63	−0.35	0.39	0.31	0.11	−0.63
NM	ATmean	0.12	0.11	0.13	0.15	0.58	−0.39	−0.51	−0.57	0.25
	ATmax	−0.34	0.00	−0.30	0.05	−0.64	0.45	0.42	0.53	−0.14
	ATmin	0.19	−0.29	0.21	−0.17	−0.34	0.31	0.17	0.69 *	−0.31
	HU	−0.50	−0.06	−0.06	−0.45	−0.86 *	0.02	0.80 *	−0.42	0.45
	PR	−0.28	0.25	0.19	−0.36	0.64	0.46	−0.48	0.44	−0.35
	WS	0.76 *	0.17	0.61	−0.52	−0.62	0.22	0.24	0.72 *	−0.10
	AP	−0.10	0.08	−0.14	0.65	0.35	0.43	−0.40	0.05	−0.06
	STmean	−0.13	0.11	0.02	−0.33	−0.48	0.21	0.45	0.26	−0.09
	SD	0.58	−0.42	−0.28	−0.45	−0.81 *	0.05	0.45	0.15	−0.18
	STmax	−0.04	−0.07	0.32	0.37	0.28	−0.16	−0.25	−0.23	0.03
	STmin	0.00	0.00	−0.45	0.29	0.05	−0.30	−0.26	−0.56	0.12

All months are denoted using the first three letters as abbreviations. The “*” mark indicates that the correlation between this meteorological factor and the $GDI$ in that region is highly significant.

). There were significant positive correlations between $GDI$ and $HU$, and $PR$, in the same season with correlation coefficients of −0.76 and −0.72. In July, they respectively reached −0.84 and −0.71. The partial correlation results showed negative correlation between $GDI$ and $PR$ in autumn, indicating that a rainy autumn is beneficial for grassland vegetation cover (Figure 5).

In the NM, the $GDI$ showed a negative significant correlation with $AP$ in spring and summer, with correlation coefficients of −0.60 and −0.67, which is impacted by the value of −0.61 in May and the value of −0.66 in June. In addition, the partial correlation results showed a significant negative correlation between $GDI$ and $HU$ in spring, with a correlation coefficient of −0.69. This illustrates an inhibiting effect of the rainy climate on grassland development in arid areas. This finding is consistent with the similar correlation between $GDI$ and $AP$ in other arid areas (NN and IM) in China. Furthermore, the partial correlation results showed significant positive correlation between $GDI$ and $WS$ in spring and autumn, with correlation coefficients of 0.80 and 0.52, especially in March (r = 0.76) and in October (r = 0.76). This indicates that a higher $WS$ tends to cause plants to fall at the beginning and end of the growing season, playing an important role in the response of vegetation growth to climate change in the NM region. Considering that loess and wind−blown sand landscapes are very common in this region, larger $WS$ will lead to an increase in the frequency of sandstorm formation, which will affect plant growth at the beginning of the growing season.

In the TP, the $GDI$ showed a significant positive correlation with atmospheric pressure in winter and summer, with correlation coefficients of 0.66 and 0.59. Subsequently, we examined the partial correlations between $GDI$ and climatic variables under different seasons in the TP region. $AP$ held the most obvious correlation with $GDI$ in winter (r = 0.82). Under global warming, the continuous increase in heterogeneity of atmospheric water vapor pressure leads to a drier atmospheric environment in the Tibetan Plateau, obviously reducing the productivity of grasslands and increasing the degree of degradation. The partial correlation results showed a negative correlation between $GDI$ and $AP$ in spring, especially in March, where the correlation coefficient value reached −0.83. This indicates that water conditions play an important role in the response of vegetation growth to climate change in the TP region. This aligns with observations of the critical role of water conditions in the responses of phenology of marsh wetlands to climate change on the Tibetan Plateau [38,43]. Therefore, compared with other factors, both water conditions and gas exchange play a significant role at high altitudes.

In the NN, the $GDI$ showed a positive correlation with $WS$ in spring and summer, with correlation coefficients of 0.73 and 0.52. However, the partial correlations between $GDI$ and $WS$ in all seasons were not statistically significant (p ≥ 0.05). This can be attributed to its location within a temperate monsoon climate region. Affected by the southeast monsoon, this region experiences not only strong winds but also high temperature and heavy rainfall, especially in spring and summer. These factors are beneficial to the growth of grassland vegetation, neutralizing the increasing grassland degeneration index caused by $WS$. In spring, the $WS$ showed a significant negative correlation (r = −0.66) with $AP$, which is impacted by the value of −0.52 in May. Meanwhile, the $GDI$ showed a negative correlation with $S{T}_{\min }$ in spring and summer (r = −0.72 and −0.58), indicating the remarkable influence of $AP$ and daily $S{T}_{\min }$ on $GDI$ in spring.

In the SM, the $GDI$ showed a significant positive correlation with wind speed in summer, with a correlation coefficient of −0.59, which is influenced by the value of −0.67 in July. Meanwhile, in spring, a significant positive partial correlation was discovered between $GDI$ and sunshine duration, whereas a significant negative partial correlation was observed between $GDI$ and minimum surface temperature. These results demonstrate that grassland vegetation growth in the SM region depends mainly on daylight and nighttime soil temperature during spring.

4. Discussion

Our results provide the first nationwide images and direct evidence of the grassland degradation index in China since the 21st century. This is consistent with previous observations of instability changes in grassland condition [4], but our study offers new findings as well. The $GDI$ showed obvious spatial variability in different ecological function regions of China, fluctuating between mild degradation and moderate degradation. The Inner Mongolia Plateau Grassland region held the highest $GDI$ and most unstable range. This finding is largely consistent with findings revealed in the 1980–2010 period in this region [16]. The overall situation of grassland degradation has improved in the recent 20 years, especially in the NN and TP. They showed a downward trend of $GDI$, indicating an improvement in grassland productivity during the study period. Additionally, there is an obvious response of the $GDI$ to climate change. In arid regions, $GDI$ is influenced by precipitation (negative) and temperature (positive). In the Tibetan Plateau, $GDI$ has an obvious relationship with wind speed.

We made substantial efforts to verify the quality of the datasets we used. Consequently, we excluded stations where changes in measurement sites and instruments occurred, and where there were large data gaps. Ultimately, only 231 stations were used and there were different amounts of data for 11 climatic factors in a given region, resulting in some gaps (Figure 1). The distribution of meteorological stations in the Qinghai–Tibetan Plateau Grassland region is uneven, potentially skewing the representation of $GDI$ trends across the entire region. Therefore, the description of the grassland situation in the TP region does not include the northwestern TP in this paper. These findings provide baseline data on grassland degradation all over China, offering references for scientific inquiry and practical applications in further studies.

At the same time, the fluctuation of $GDI$ in the IM region has aroused our attention. In a region with animal husbandry as the main economic source, excessive grazing is a factor among the human factors that has a significant impact on $GDI$ changes. However, considering that it is difficult to obtain intuitive data from this perspective, we have investigated the grazing policies in the IM. The results show that only a few policies have implemented grazing prohibition measures in the 19 years of the study period; all of them are short−term grazing bans to restore the ecology of the grassland, usually no more than 3 months, and there are no related policies to stop grazing. Therefore, the effect of this view on the resulting data is not obvious.

Further exploration is needed to incorporate human interference, such as urbanization processes and nighttime light indices, in order to formulate strategies for grassland conservation and sustainable utilization according to local grassland degradation conditions.

5. Conclusions

Our study reveals several crucial findings regarding $GDI$ in five ecological function regions across China. First, we observed a significant spatial variability during the study period. Second, the annual change of $GDI$ implied that the grassland degradation condition has not continued to worsen since the beginning of 21st century; instead, there are clear trends of grassland improvement in some regions. Notably, there was a restoration of grassland at the rate of 0.17 ($GDI$)/decade in IM.

Furthermore, our evidence shows that precipitation, temperature, wind, and atmospheric pressure, which influence grassland growth and development in different ecological faction regions, play a regulatory role in the impact of climate change on the $GDI$. Water condition plays an important role in grassland growth in most of the regions, except for the SM with its wet and rainy climate. Meanwhile, the negative correlations are much more obvious in spring and summer, which is the time before the plant matures. This influence is particularly prominent in IM and NM throughout the whole year. Temperature has a more significant partial correlation with $GDI$, and this phenomenon is more obvious in SM, NM, and IM. In addition, for different regions, the partial correlation between temperature and $GDI$ is often very different. For example, the partial correlation between ground temperature and $GDI$ is significantly positive in the SM in winter (r = 0.64), while it is completely opposite in NM (r = −0.75). The effect of wind speed on $GDI$ is mainly reflected in NM, NN, and SM, and it inhibited the $GDI$ of SM at the annual level. Among them, the $GDI$ showed a positive partial correlation with wind speed in spring. This is because sandstorms with higher wind speeds can bury plants that are just beginning to grow in the spring, leading to degradation of the grassland. At the same time, in the SM, the $GDI$ showed a positive partial correlation with wind speed in winter. This indicates that in the warm climate of SM, most vegetation will not wither in winter, so it easily collapses due to high wind speeds when growing to a certain height. Finally, as a region with a plateau mountain climate, the average altitude of TP is over 4000 m. Therefore, $AP$ continues to have a negative effect on plant growth in this region, especially in winter.

This study highlights the influences of water condition and wind on the $GDI$ of grassland vegetation, particularly in the context of global warming. It underscores the importance of considering hydrology supply and airflow transition in vegetation simulations conducted by terrestrial ecosystem models in cold and dry regions worldwide.