An Approach to the Temporal and Spatial Characteristics of Vegetation in the Growing Season in Western China

: Since the implementation of the great western development strategy in 2000, the ecological environment in the western region of China has been signiﬁcantly improved. In order to explore the temporal and spatial characteristics of vegetation coverage in the western region, this paper adopted the method of Maximum Value Composite (MVC) to obtain the mean Normalized Di ﬀ erence Vegetation Index (NDVI) of vegetation on the basis of the Moderate-resolution Imaging Spector audiometer (MODIS) data of 2000 / 2005 / 2010 / 2015 / 2018. Thereafter, the spatio-temporal di ﬀ erentiation characteristics of vegetation in western China were analyzed. The results show that: (1) According to the time characteristics of vegetation coverage in the western region, the average annual NDVI value of vegetation coverage in the growing season in the western region ﬂuctuated between 0.12 and 0.15, among which that of 2000 to 2010 ﬂuctuated more greatly but did not show obvious change trend. (2) Based on Sen trend and Mann-Kendall test analysis, the area of vegetation coverage improvement in the western region from 2000 to 2018 was larger than that of signiﬁcant vegetation degradation. (3) From the perspective of global autocorrelation coe ﬃ cient, Moran’s I values were all positive from 2000 to 2018, which indicates that the vegetation coverage in the west showed strong positive autocorrelation in each period. According to the average value and coe ﬃ cient of variation of vegetation coverage, the vegetation coverage was lower in 2000, its internal variation was smaller, and the vegetation coverage increased with time. According to the local spatial autocorrelation analysis, the vegetation coverage levels in di ﬀ erent regions varied greatly. (4) The standard deviation ellipse method was used to study the spatial distribution and directional transformation of vegetation. It makes the result more intuitive, and the three levels of gravity center shift, direction shift, and angle shift were considered: the vegetation growth condition in the spatial aggregation area improved in 2015; the standard deviation ellipses in 2000 and 2018 overlapped and shifted eastward, which indicates that the vegetation coverage conditions in the two years were similar and got ameliorated.


Introduction
In the international approaches to vegetation degradation, generally Remote Sensing (RS), Global Position System (GPS), and Geographic Information System (GIS) are combined to monitor vegetation degradation. For example, Mustafa [1] studied the evaluation method of vegetation degradation in abandoned coal mines using spatial information technology and analyzed the vegetation degradation status before and after coal mining. Anilz Chitade [2] et al. analyzed the land use/cover changes of underdeveloped area in China and needs to be further developed. Since the reform and opening, with the scale of mineral resources development being continuously increased in the 12 western provinces (Figure 1), they have increasingly become the main area of mineral resources development in China. At the end of 2010, the country's coal reserves were ascertained to be 1341.2 billion tons, an increase of about 300 billion tons over 2005, with the western region accounting for more than 90% of the national increase [22,23]. According to the following pictures of the typical surveyed areas (Figure 2), as the coal resources in the central and eastern regions are drying up, coal development is accelerating to the west, where the ecological environment is fragile. Owing to the fragile ecological environment in the west, it faces the problem of relative imbalance between development and protection. Therefore, the vegetation coverage in west China is more susceptible to external disturbances. Analysis of the spatio-temporal variation characteristics of vegetation in this region is of great significance for evaluating its ecological environment and implementing reasonable control measures.

Data Sources and Processing
Since Normalized Difference Vegetation Index (NDVI) and EVI data developed by MODIS have different descriptions of high-density vegetation in humid environment, this feature of EVI data provides a new way to study the seasonal changes of high-coverage vegetation. EVI data is more closely related to the monthly average temperature, so EVI data can be used to study the relationship between vegetation and temperature in the later period. In this study, NDVI was selected as the data source for the western region of China, which have a similar ability to describe low-cover vegetation in the arid and sub-humid environment.

Data Sources and Processing
Since Normalized Difference Vegetation Index (NDVI) and EVI data developed by MODIS have different descriptions of high-density vegetation in humid environment, this feature of EVI data provides a new way to study the seasonal changes of high-coverage vegetation. EVI data is more closely related to the monthly average temperature, so EVI data can be used to study the relationship between vegetation and temperature in the later period. In this study, NDVI was selected as the data Remote Sens. 2020, 12, 945 5 of 16 source for the western region of China, which have a similar ability to describe low-cover vegetation in the arid and sub-humid environment.
The remote sensing data are from the MOD13Q1 data of the vegetation growing season from May to September (from 129 days to 273 days), the 800 scenes, the 19th Issue, 2000-2018, United States Geological Survey (USGS) (https://search.earthdata.nasa.gov/search) with a spatial resolution of 250m and a temporal resolution of 16 days. Firstly, MODIS Reprojection Tool (MRT) software was used to convert the data format and projection, then MODIS NDVI image data for 2000/2005/2010/2015/2018 were obtained by cutting out the boundaries of 12 provinces in the west, and then The Environment for Visualizing Images (ENVI5.3) software was used to synthesize annual NDVI 1 data and monthly NDVI 1 data by Maximum Value Composite (MVC) to remove the influence of cloud, fog and solar altitude angle as much as possible. Since the annual maximum NDVI can effectively reflect the regional vegetation coverage, and the time point when the vegetation flourish is not easy to determine, the annual maximum NDVI 1 of each pixel was selected as the research data. Finally, the Band Math operation expression (b1 It 0) * 0 + (b1 ge 0) * (b1 * 0.0001) was used to make NDVI 1 value between 0 and 1. The Band Math operation expression of vegetation coverage((b1 lt 0.031766) * 0 + (b1 gt 0.522991) * 1 + (b1ge 0.031766 and b1 le 0.522991) * ((b1 − 0.031766)/(0.522991 − 0.031766) and the calculation method of annual maximum mean are as follows: In the formula, Pi is the average of the monthly maximum NDVI of the pixels from May to September, n is the number of regional pixels, verage is the normalized coefficient of vegetation coverage, and the reference value is 0.0121165124 [24,25].

Sen Trend Analysis and Mann-Kendall Test
The non-parametric Sen trend method was used to calculate the change trend of vegetation coverage in west China. The advantage of Sen trend that it has no requirement on data distribution makes it suitable for analysis of missing data, which could reduce the interference of abnormal values, and get SEN by calculating the median value of the sequence [26].
β= Median In the formula, i and j represent year i and year j, respectively, Xi and Xj represent year i NDVI 1 and year j NDVI 1 , respectively, β represents trend degree, and the value of β indicates the rise and fall of the vegetation trend. When β > 0, the time sequence shows an upward trend, while when β < 0, the time sequence shows a downward trend.
However, Sen cannot judge the significance of sequence trend by itself. Mann-Kendall statistical test was applied to test the significance of Sen trend. Mann-Kendall is a nonparametric statistical method [27,28]. The process of the trend checking method is as follows: for the sequence X = (x 1 , x 2 , . . . , x n ), the magnitude relation between Xi and Xj (set as s) in all dual values (between Xi and Xj, j > i) is first determined. The following assumptions are made: sgn0: the data in the sequence are randomly arranged, i.e., there is no significant trend; sgn1: the sequence has a monotone trend of increasing or decreasing. The inspection statistic S is calculated by Formula (3): Statistic S is defined as follows: Remote Sens. 2020, 12, 945 In selecting inspection statistics, when n < 10, the statistic s is directly used for bilateral trend inspection; when n ≥ 10, the statistic s approximately follows the standard normal distribution, then z is used for trend inspection, and the z value is calculated by Formula (5): In the formula, z is the standardized test statistic, the value range of statistic z is (−∞, +∞), S is the test statistic, n is the number of sequence samples, m is the number of knots (repeated data sets) in the sequence, t i is the width of the knots (the number of repeated data in group I of repeated data sets). Xi and Xj respectively represent the NDVI value in year i and year j. Given the significance level α, the critical value Z1-α/2 was found in the normal distribution table; when |Z| ≤ Z1α/2, the original assumption is accepted, that is, the trend is not significant; when |Z| > Z1α/2, the original assumption is rejected, i.e., the trend is significant. Then Z1α/2 = 1.96. When β > 0 and Z1α/2 > 1.96, the sequence shows a significant upward trend; when β > 0 and |Z| ≤ Z1α/2, the sequence shows an upward but insignificant trend; similarly, when β < 0 and Z1α/2 > 1.96, the sequence shows a significant downward trend; when β < 0 and |Z| ≤ Z1α/2, the sequence shows a downward but insignificant trend.
In this paper, the trend inspection was carried out by using the inspection statistic S. Given the significance level α, if |S| ≥ S α/2, W0 is not permitted to determine that the original sequence has a significant trend, otherwise W0 is permitted to determine that the sequence trend is not significant. If S > 0, the sequence is considered to have an upward trend; if S = 0, there is no trend; if S < 0, the sequence is considered to have a downward trend.

Spatial Autocorrelation
Spatial autocorrelation analysis can reveal the correlation between a certain regional unit and neighboring regional units in terms of common geographic features or attribute values [29,30] and can be analyzed from global and local perspectives. The commonly used spatial autocorrelation measure indexes are Moran's I value and Local indication of spatial association (LISA) value. Global Moran's I can reveal the global spatial correlation of vegetation coverage. Local Moran's I index can reveal the local spatial aggregation characteristics of vegetation coverage in west China. Its essence is that the index is high, indicating similar spatial unit aggregation; otherwise, it indicates different spatial unit aggregation [31,32]. The calculation formula is as follows: In the formula, n is the number of spatial units in the research area; X i and X j are the vegetation coverage (attribute values) of spatial units i and j respectively, X is the average value of all the attribute values mentioned above; W ij is the spatial weight: S 2 is the variance of vegetation coverage of each spatial unit (8); I is global Moran's I and I i is local LISA index. If I > 0, it is positive correlation; If I < 0, it is negative correlation; if the absolute value of I tends to be 1, there is strong spatial autocorrelation.

Standard Deviation Ellipse
Standard deviation ellipse (SDE) can reveal geographical spatial characteristics and spatio-temporal evolution from the perspectives of center of gravity, distribution range, direction and shape, etc. [33]. It is a method to accurately reveal various economic distribution characteristics by using spatial statistical methods. Its shape index (the ratio of the length of the minor axis to the major axis of the ellipse) reflects the shape of the distribution of elements. When the value is between 0 and 1, the closer it is to 1, the closer the shape of the main area where the spatial distribution of elements is to the standard circle, otherwise the closer it is to 0, the closer the shape is to linearity.
X and Y respectively represent the longitude and latitude coordinates of the vegetation coverage center; X i and Y i respectively correspond to the longitude and latitude coordinates of vegetation coverage on grid i, and M i represents the value of vegetation coverage grid i. X-axis length and y-axis length of elliptic distribution: Aazimuthal angle:

Temporal Variation Characteristics of Normalized Difference Vegetation Index (NDVI) in Growing Season of West China
In order to study the characteristics of NDVI changes with time in the growing season in the western region [34], the average NDVI values of the vegetation coverage areas in the annual NDVI images from 2000 to 2018 were taken to represent the vegetation coverage status in that year, and an inter annual NDVI change figure was made ( Figure 3). As can be seen in Figure 3, the average annual

An Analysis of the Spatial Variation Trend of NDVI in Growing Season in West China
The combination of Theil-Sen Median trend analysis and Mann-Kendall test [35] can effectively reflect the spatial distribution characteristics of vegetation NDVI change trend during the growing season in west China from 2000 to 2018. From Table 1, it can be seen that the vegetation activities in the western region were obviously degraded on the whole, and some areas were also improved. The vegetation degradation was obvious in 4.63% of the regions, the vegetation of 65.91% of the regions was degraded and relatively stable, the vegetation of 29.30% of the regions was improved, and the vegetation in 0.16% of the area has a significant improvement trend. From the above data, it can be seen that the change trend of vegetation NDVI in the growing season in the west showed a significant degradation trend, and some areas also began to develop towards an upward trend. Based on Sen's trend analysis and Mann-Kendall test (M-Ktest), the change trend of NDVI on pixel scale was analyzed ( Figure 4). According to the number of pixels, the degraded and stable regions account for the largest proportion, the significantly improved regions account for a smaller proportion, and the significantly degraded regions account for a smaller proportion than the slightly improved regions, indicating that vegetation degradation has ameliorated.
As for the distribution area of vegetation degradation, there are many degraded pixels in northwest Qinghai, northwest Xinjiang, most of Tibet, and northern Sichuan; the regions with improved NDVI in the growing season in west China are mainly distributed in the southwest of Gansu, the central and southern regions of northern Shaanxi, the eastern region of Inner Mongolia, the northeast of Qinghai, the western part of Xinjiang, the southern part of Sichuan, the northern and eastern parts of Yunnan, the southern parts of Guizhou and Guangxi, and Chongqing; the eastern part of Xinjiang, the western part of Tibet, the northwestern part of Gansu, the western part of Qinghai, and the western part of eastern Mongolia showed a trend of degeneration, though were with little change and were relatively stable. According to the analysis of the above phenomena, the

An Analysis of the Spatial Variation Trend of NDVI in Growing Season in West China
The combination of Theil-Sen Median trend analysis and Mann-Kendall test [35] can effectively reflect the spatial distribution characteristics of vegetation NDVI change trend during the growing season in west China from 2000 to 2018. From Table 1, it can be seen that the vegetation activities in the western region were obviously degraded on the whole, and some areas were also improved. The vegetation degradation was obvious in 4.63% of the regions, the vegetation of 65.91% of the regions was degraded and relatively stable, the vegetation of 29.30% of the regions was improved, and the vegetation in 0.16% of the area has a significant improvement trend. From the above data, it can be seen that the change trend of vegetation NDVI in the growing season in the west showed a significant degradation trend, and some areas also began to develop towards an upward trend. Based on Sen's trend analysis and Mann-Kendall test (M-Ktest), the change trend of NDVI on pixel scale was analyzed (Figure 4). According to the number of pixels, the degraded and stable regions account for the largest proportion, the significantly improved regions account for a smaller proportion, and the significantly degraded regions account for a smaller proportion than the slightly improved regions, indicating that vegetation degradation has ameliorated.
indicates that vegetation was affected by natural factors, such as temperature, precipitation and landform. The frequent occurrence of landslides and debris flows in the rainy season in the southwest made vegetation show degraded trend. Due to the implementation of the National Natural Forest Protection Project in recent years, the ecological environment has been improved to a great extent and the vegetation degradation has been improved. However, with the acceleration of urbanization, the rapid increase of population, and the need to improve infrastructure and management in the west, the restoration of vegetation cover has also been affected.   Table 2: Moran' s I > 0 indicates a positive spatial correlation. The greater the value, the more obvious the spatial correlation is. Moran's I < 0 indicates a negative correlation in the spatial region. The smaller the value, the greater the spatial difference is. Otherwise, if Moran's I = 0, the space is at random. The aggregation characteristics of different administrative regions in the whole western vegetation coverage are available from the table. As far as the spatial autocorrelation coefficient of vegetation coverage is concerned, from 2000 to 2018, Moran's I values were all positive numbers, the significance level of which can be determined by the P-value test of standardized Z values. Then comparing it with the significance level (α = 0.05) for it to pass the significance test (P < 0.01), rejecting the zero hypothesis, the research revealed that the vegetation coverage in west China showed strong As for the distribution area of vegetation degradation, there are many degraded pixels in northwest Qinghai, northwest Xinjiang, most of Tibet, and northern Sichuan; the regions with improved NDVI in the growing season in west China are mainly distributed in the southwest of Gansu, the central and southern regions of northern Shaanxi, the eastern region of Inner Mongolia, the northeast of Qinghai, the western part of Xinjiang, the southern part of Sichuan, the northern and eastern parts of Yunnan, the southern parts of Guizhou and Guangxi, and Chongqing; the eastern part of Xinjiang, the western part of Tibet, the northwestern part of Gansu, the western part of Qinghai, and the western part of eastern Mongolia showed a trend of degeneration, though were with little change and were relatively stable. According to the analysis of the above phenomena, the vegetation in the southwest and the south was in good condition and showed degraded pixels, which indicates that vegetation was affected by natural factors, such as temperature, precipitation and landform. The frequent occurrence of landslides and debris flows in the rainy season in the southwest made vegetation show degraded trend. Due to the implementation of the National Natural Forest Protection Project in recent years, the ecological environment has been improved to a great extent and the vegetation degradation has been improved. However, with the acceleration of urbanization, the rapid increase of population, and the need to improve infrastructure and management in the west, the restoration of vegetation cover has also been affected.  Table 2: Moran' s I > 0 indicates a positive spatial correlation. The greater the value, the more obvious the spatial correlation is. Moran's I < 0 indicates a negative correlation in the spatial region. The smaller the value, the greater the spatial difference is. Otherwise, if Moran's I = 0, the space is at random. The aggregation characteristics of different administrative regions in the whole western vegetation coverage are available from the table. As far as the spatial autocorrelation coefficient of vegetation coverage is concerned, from 2000 to 2018, Moran's I values were all positive numbers, the significance level of which can be determined by the p-value test of standardized Z values. Then comparing it with the significance level (α = 0.05) for it to pass the significance test (p < 0.01), rejecting the zero hypothesis, the research revealed that the vegetation coverage in west China showed strong positive autocorrelation in various periods [36]. That is to say, the vegetation coverage in west (2) Local aggregation feature analysis The changes of vegetation coverage from the previous checking point in the western region in 2000, 2005, 2010, 2015, and 2018 were calculated respectively, and the correlation analysis of spatial local aggregation pattern was carried out. The spatial local indication of spatial association (LISA) cluster diagram of vegetation coverage changes in the western region was drawn using ArcGIS [40,41]: (Figure 5). The discussion of the evolution characteristics of local aggregations came to the conclusion that the vegetation coverage levels in the western, southwestern and northwestern provinces were quite different. Among them, the western regions such as Xinjiang, Gansu, Qinghai, and Tibet were The discussion of the evolution characteristics of local aggregations came to the conclusion that the vegetation coverage levels in the western, southwestern and northwestern provinces were quite different. Among them, the western regions such as Xinjiang, Gansu, Qinghai, and Tibet were characterized by L-L aggregation; the southwest regions such as Sichuan, Chongqing, Guizhou, Yunnan, and Guangxi showed H-H aggregation characteristics; the northern Inner Mongolia region remained in an insignificant type for 5 years. Based on Sen's trend analysis and Mann-Kendall test (M-K test), the change trend of NDVI on pixel scale was analyzed (Figure 4). According to the number of pixels, the degraded and stable regions account for the largest proportion, the significantly improved regions account for a smaller proportion, and the significantly degraded regions account for a smaller proportion than the slightly improved regions, indicating that vegetation degradation has ameliorated [42,43]; from 2000 to 2018, the vegetation coverage characteristics in Gansu Province changed from L-L aggregation to being insignificant, indicating that the vegetation degradation in this region was relatively significant; from 2000 to 2010, Xinjiang, Gansu, and Qinghai were transformed into insignificant types, indicating that their local spatial autocorrelation was weakened and vegetation degradation was manifested [44]; from 2000 to 2018, Xinjiang and Qinghai kept characterized by L-L aggregation, which indicates that their vegetation coverage was relatively stable; from 2000 to 2018, Chongqing and Guizhou had been characterized by H-H aggregation, which indicates that their vegetation growth was relatively good.

Spatiotemporal Evolution Characteristics of Annual Vegetation Coverage Based on Standard Deviation Ellipse (SDE)
Based on the calculation of vegetation coverage (FVC), the spatial distribution pattern of vegetation in the growing period of west China was analyzed by standard deviation ellipse (SDE) [45]. In this study, SDE was used to analyze the monthly mean value of NDVI of vegetation in the growth period of west China, and ArcGIS SDE tool was used to calculate the central value of the mean value of vegetation in the growth period of west China ( Figure 6). transformed into insignificant types, indicating that their local spatial autocorrelation was weakened and vegetation degradation was manifested [44]; from 2000 to 2018, Xinjiang and Qinghai kept characterized by L-L aggregation, which indicates that their vegetation coverage was relatively stable; from 2000 to 2018, Chongqing and Guizhou had been characterized by H-H aggregation, which indicates that their vegetation growth was relatively good.

Spatiotemporal Evolution Characteristics of Annual Vegetation Coverage Based on Standard Deviation Ellipse (SDE)
Based on the calculation of vegetation coverage (FVC), the spatial distribution pattern of vegetation in the growing period of west China was analyzed by standard deviation ellipse (SDE) [45]. In this study, SDE was used to analyze the monthly mean value of NDVI of vegetation in the growth period of west China, and ArcGIS SDE tool was used to calculate the central value of the mean value of vegetation in the growth period of west China ( Figure 6). From the coordinates of the center point in Table 3, it can be seen that the average annual centers of gravity of the western vegetation were all located in Qinghai Province. Figure 6 shows the transfer of the total spatial distribution pattern of standard deviation ellipse (SDE) center point [46,47]. The sampling points were evenly distributed in the western provinces, and the areas covered by ellipses included western Inner Mongolia, eastern Xinjiang, northeastern Tibet, northern Yunnan, northern Guizhou, Sichuan, Gansu, Chongqing, Qinghai, Ningxia, and Shaanxi. With the change of time, due to the difference of vegetation growth status, the center of gravity of vegetation in the growth period of west China has shifted and the angle of change has varied (Table 3), making the spatial distribution From the coordinates of the center point in Table 3, it can be seen that the average annual centers of gravity of the western vegetation were all located in Qinghai Province. Figure 6 shows the transfer of the total spatial distribution pattern of standard deviation ellipse (SDE) center point [46,47]. The sampling points were evenly distributed in the western provinces, and the areas covered by ellipses included western Inner Mongolia, eastern Xinjiang, northeastern Tibet, northern Yunnan, northern Guizhou, Sichuan, Gansu, Chongqing, Qinghai, Ningxia, and Shaanxi. With the change of time, due to the difference of vegetation growth status, the center of gravity of vegetation in the growth period of west China has shifted and the angle of change has varied (Table 3), making the spatial distribution of standard deviation ellipse present a "northeast-southwest" pattern. The shape index in 2010 was 0.642, which shows that the shape of the main area where the spatial distribution of vegetation coverage was located in the vegetation growth season in west China in 2010 was relatively close to the standard circle. Let analysis and discussion refer to the transfer direction of the above ellipse: from 2000 to 2005, the standard deviation ellipse shifted westward, which was consistent with the aggregation type of LISA cluster diagram, indicating that under the natural and artificial influence, vegetation degradation appeared [48]; from 2005 to 2010, the standard deviation ellipse shifted eastward, which was consistent with the spatial autocorrelation distribution of vegetation coverage, indicating that vegetation growth was improved by human and self-repairing measures by 2010; from 2010 to 2015, the standard deviation ellipse shifted to the west, and from 2015 to 2018, the standard deviation ellipse shifted to the east. The directions of the ellipses in 2015 and 2018 were the same, and the center of gravity of the ellipse shifted eastward in 2018, which indicates that the vegetation growth had improved on the basis of the spatial aggregation region in 2015. The standard deviation ellipses in 2000 and 2018 partially overlapped and shifted to the east, which indicates that the vegetation coverage in these two years was relatively good and the vegetation growth was relatively consistent, It is proved that compared with the standard deviation ellipse method, which is used to research in meteorological station and economics, etc. [49,50]. In this study, it is of great significance to study the spatial change of vegetation cover, as it makes the results more intuitive and concrete.

Conclusions
(1) From the perspective of the spatial pattern of vegetation coverage in the growing season in west China, the proportion of NDVI and the distribution area of vegetation were analyzed and discussed. The areas with significant degradation distribution included the west adjacent to the east of Inner Mongolia, the northwest of Tibet, the northwest of Qinghai Province, a small number of areas in the east and north of Xinjiang, central Ningxia, the northwest of Gansu Province, while the areas with good vegetation coverage and significant improvement were Yunnan, Guizhou, Sichuan, Chongqing, Guangxi, the southeast of Gansu, the north of Ningxia, and the east of Inner Mongolia.
(2) According to the analysis of vegetation change trend in growing season in west China and NDVI change trend, from 2000 to 2018, the area with improved surface vegetation coverage was larger than the area with obvious vegetation degradation. The improved area accounted for 29.46% of the vegetation coverage area, the significantly degraded area accounted for 4.63%, and 65.91% of the areas were degraded and relatively stable.
(3) From the analysis of spatial autocorrelation changes of vegetation in the growing season in west China, the global spatial autocorrelation analysis shows that the vegetation coverage was lower in 2000, but the vegetation coverage increased with the passage of time, the vegetation growth status ameliorated, and its internal variation was relatively small. The autocorrelation analysis of local aggregation space shows that the vegetation coverage levels in the western, southwestern and northwestern provinces were quite different. Among them, the western regions such as Xinjiang, Gansu, Qinghai and Tibet were characterized by L-L aggregation; the southwest regions such as Sichuan, Chongqing, Guizhou, Yunnan, and Guangxi showed H-H aggregation characteristics; the northern Inner Mongolia region remained an insignificant type for 5 years.
(4) Based on the analysis of the spatial pattern change of the standard deviation ellipse of vegetation in the growing season in west China, the shift of center of gravity, direction, and angle of the standard deviation ellipse showed that the vegetation growth has improved on the basis of the spatial aggregation region in 2015. The standard deviation ellipses in 2000 and 2018 partially overlapped and shifted to the east, which indicates that the vegetation coverage in these two years was relatively good and the vegetation growth was relatively consistent. (5) In future studies, the spatio-temporal differentiation of the driving mechanism and factors will be analyzed [51,52]. It is of significance for vegetation restoration and protection and vegetation restoration in West China.