Greening Implication Inferred from Vegetation Dynamics Interacted with Climate Change and Human Activities over the Southeast Qinghai–Tibet Plateau

Vegetation dynamics are sensitive to climate change and human activities, as vegetation interacts with the hydrosphere, atmosphere, and biosphere. The Yarlung Zangbo River (YZR) basin, with the vulnerable ecological environment, has experienced a series of natural disasters since the new millennium. Therefore, in this study, the vegetation dynamic variations and their associated responses to environmental changes in the YZR basin were investigated based on Normalized Difference Vegetation Index (NDVI) and Global Land Data Assimilation System (GLDAS) data from 2000 to 2016. Results showed that (1) the YZR basin showed an obvious vegetation greening process with a significant increase of the growing season NDVI (Zc = 2.31, p < 0.05), which was mainly attributed to the wide greening tendency of the downstream region that accounted for over 50% area of the YZR basin. (2) Regions with significant greening accounted for 25.4% of the basin and were mainly concentrated in the Nyang River and Parlung Tsangpo River sub-basins. On the contrary, the browning regions accounted for <25% of the basin and were mostly distributed in the urbanized cities of the midstream, implying a significant influence of human activities on vegetation greening. (3) The elevation dependency of the vegetation in the YZR basin was significant, showing that the vegetation of the low-altitude regions was better than that of the high-altitude regions. The greening rate exhibited a significantly more complicated relationship with the elevation, which increased with elevated altitude (above 3500 m) and decreased with elevated altitude (below 3500 m). (4) Significantly positive correlations between the growing season NDVI and surface air temperature were detected, which were mainly distributed in the snow-dominated sub-basins, indicating that glaciers and snow melting processes induced by global warming play an important role in vegetation growth. Although basin-wide non-significant negative correlations were found between precipitation and growing season NDVI, positive influences of precipitation on vegetation greening occurred in the arid and semi-arid upstream region. These findings could provide important information for ecological environment protection in the YZR basin and other high mountain regions.


Introduction
Vegetation is an important component in terrestrial ecosystems that not only reflects the land surface condition but also interacts with the hydrosphere, atmosphere and biosphere [1]. Therefore, vegetation is often regarded as a good indicator of environmental changes [2] and can be used to describe interactive impacts between environmental changes and terrestrial ecosystems [3]. In recent years, spatial and temporal variations of vegetation have attracted considerable attention and persistent interest globally. Zhu et al. used the Leaf Area Index (LAI) to study vegetation dynamics and found that the globe experienced an obviously greening process, with CO 2 fertilization effects playing an important role in this greening trend [4]. Greening trends mainly occurred in Europe [5], eastern North America [6], northern Africa [7,8], and China [9]. Chen et al. further indicated that China was more responsible for "Greening Earth" due to the implementation of environmental protection projects since the new millennium [10]. Lu et al. also considered that human activities in China were the most significant factors for vegetation dynamics, and the implementation of large-scale natural conservation measurements was more effective for vegetation recovery [11]. Hua et al. pointed out that land-use and land-cover change determined the vegetation greening process in China, and North China was clearly becoming green because of the expansion in agricultural land, forest, and grasslands [12]. Many studies have also concluded that many regions in China exhibited a greening process, such as the Loess Plateau [13], Northwest inland [14], and Southwest China [15]. However, the increasing sensitivity of vegetation to climate change and human activities along an elevational gradient in high-altitude regions raises a great challenge to the understanding of the interaction between vegetation dynamics and environmental changes.
Global climate change, mainly characterized by global warming, has altered the water and energy balance of terrestrial ecosystems by redistributing climate factors in space and time, which will significantly affect vegetation spatiotemporal variations [16][17][18]. As the two main climate factors, precipitation and surface air temperature directly drive the water and energy cycle and further affect the spatiotemporal distributions of vegetation [19][20][21]. However, the vegetation response to precipitation is of great spatial heterogeneity. There generally exists a positive relationship between vegetation growth and precipitation. For example, more precipitation will promote vegetation recovery, which is particularly apparent in arid and semi-arid regions where available moisture is the key material [22,23], whereas precipitation exerts negative influences on vegetation growth in some regions. Wu et al. focused on the Guangdong Province of South China, located in a tropical and subtropical region, and concluded that vegetation greening had a negative relationship with precipitation amount because of extreme precipitation events, such as freezing rain and ice storms [24]. A negative correlation between precipitation and vegetation cover was also found in Nepal [25] and the Three-River Headwaters region of the Qinghai-Tibet Plateau, which is the source of the Yangtze River, the Yellow River, and the Lantsang River [26]. Similarly, the relationship between vegetation growth and surface air temperature exhibits great geographical complexity and uncertainty. Increased surface air temperature has lengthened the growing season in northern high latitudes, which further strengthens vegetation activity and promotes vegetation growth [27,28]. However, due to continuous global warming, the surface air temperature may no longer promote vegetation growth and conversely may be detrimental to vegetation greening owing to warming-induced drought. Some studies have indicated that drought has resulted in the decrease of net primary production (NPP) in Northeast China [29], reduction of water use efficiency (WUE) in central and southern China [30], and decline of gross primary production (GPP) in northern China [31]. Soil moisture exerts an important and direct influence on vegetation growth in that it constrains evapotranspiration and photosynthesis and also maintains extra water from precipitation and snowfall. Soil moisture is also affected by climate change and simultaneously involved in a number of feedback loops at the local, regional, and global scales [32,33], and the effects of soil moisture on vegetation growth have also been analyzed. Unequivocally, great spatial heterogeneity exists in vegetation responses to precipitation and surface air temperature, which is especially apparent in the high-altitude regions with significant sensitivity and vulnerability to environmental changes [34][35][36].
The Yarlung Zangbo River (YZR) basin, whose average elevation is more than 4000 m, located in the southeast of the Qinghai-Tibet Plateau (QTP), is a typical high-altitude region. It contains many moisture transportation paths and thus is important for precipitation generation over the QTP [37]. The YZR basin is of overwhelming importance to agricultural development, sustainable environment, and economic prosperity of the QTP because 62.98% of the agricultural land and 51.30% of the population of the Tibet Autonomous Region are predominantly concentrated in the river valley of the midstream YZR. On the other hand, the effects of global warming are more evident in the YZR basin, resulting in a series of environmental changes, such as permafrost collapse [38], grassland degradation [39], and desertification [40]. In particular, drought has been a severe natural disaster in this region. You et al. pointed out that surface air temperature showed a significant upward trend with rates of 0.28 • C/decade from 1961 to 2005 over the YZR Basin [41]. Li et al. demonstrated that drought duration and magnitude in the YZR basin were becoming more severe [42]. An obvious climate transition from wet to dry in the year of 2000 was also detected by Liu et al. [43], indicating that the YZR basin has faced a more challenging environment since the new millennium. The precipitation and surface air temperature from Global Land Data Assimilation System (GLDAS) Noah model showed that the YZR basin had experienced a drying and warming process since 2000 (Supplementary Section 1). More frequent and severe droughts will pose a threat to regional water resources safety. Lu et al. pointed out that the alpine grassland cover on the QTP increased significantly from 1982 to 2013 [44]. Wang et al. considered that surface air temperature contributed significantly to vegetation cover in the QTP region [45]. Regarding the YZR basin, Han et al. also found that there was an obvious greening trend in the southeastern part of the QTP region [46]. Most previous studies used limited observed data from sparse meteorological stations to investigate the relationship between climate change and dynamic variations of vegetation [47][48][49]. There are few comprehensive studies on the spatial interaction of vegetation dynamic variations and climate change in the YZR basin due to the scarcity of in situ measurements and inadequate and irregular observation networks. Furthermore, the influence of human activities in the YZR basin cannot be ignored since there are important population centers and grain-producing areas in the middle region, where agricultural land and population is concentrated along the river valley of the midstream. Therefore, investigating the spatial responses of vegetation in the high-altitude regions to climate change and human activities is of great necessity.
The Normalized Difference Vegetation Index (NDVI), defined as the difference of red and near-infrared spectrum reflectance, reflects vegetation dynamic variations in the aspect of ecological monitoring and climate change [50]. MODIS NDVI products have been widely applied to study spatial and temporal variations of vegetation cover and their responses to global climate change in different regions, such as North America [51], South Asia [52], South America [53], Africa [54], and Europe [55]. Hence, the MOD13A3 NDVI product was used to indicate vegetation dynamics in this study. In addition, the area of the YZR basin is over 2.4 × 10 5 km 2 , while national meteorological stations are extremely scarce and climate observation networks are inadequate and irregular [56]. Thus, there are great challenges in investigating spatial and temporal variations at the basin scale. Liu et al. has validated the applicability of GLDAS NOAH data for the YZR basin using 20 hydrometeorological stations covering the upstream, midstream, and downstream regions and concluded that the spatiotemporal patterns of the GLDAS NOAH precipitation and surface air temperature data fit well with those of the observed data in this region [43]. Therefore, GLDAS NOAH data were applied to describe the spatiotemporal patterns of precipitation and surface air temperature in this study.
As mentioned above, drought in the YZR basin has become more frequent and severe since the new millennium, threatening regional water resources safety. Thus, it is imperative to identify the vegetation variations at both spatial and temporal scales and their associated responses to climate change from 2000 to 2016 in the YZR basin. Therefore, the primary objectives of this study are (1) to identify whether the vegetation in the YZR basin is greening or browning since 2000, (2) to investigate the elevation dependency of vegetation in the YZR basin with complex climate and terrain, and (3) to quantify the vegetation dynamic responses to climate change and human activities. The results will promote our understanding of vegetation dynamic variations in high-altitude regions and be useful for regional sustainable development.

Study Area
The YZR stretches from west to east through the southern edge of the QTP, with a length of about 2057 km (Figure 1a). It springs from the northwestern part of the YZR basin and is surrounded by lots of mountains, such as Tanggula Shan, Nyainqentanglha Shan, Gangdise Shan, and the Himalayas (Figure 1b). Moreover, there is a significant elevation range, from 133 m to over 7000 m, in the YZR basin. In this study, Nugesha and Nuxia hydrological stations were taken as the outlets of the upstream and midstream, respectively, which divided the YZR basin into three sub-catchments. The region between the Chemayungdung Glacier and Nugesha hydrological station is defined as the upstream and is an arid and semi-arid region with annual mean precipitation of less than 300 mm. The region between the Nugesha and Nuxia hydrological stations is defined as the midstream, which has a temperate climate. The region below the Nuxia hydrological station, with an annual mean precipitation of more than 2000 mm, is defined as the downstream characterized by a warm and wet climate. The elevation increases from the southeastern part (the downstream) to northwestern part (the upstream). In addition, there are five tributaries-Dogxung Zangbo River and Nyang Qu River (upstream), Lhasa River and Nyang River (the east of the midstream), and Parlung Tsangpo River (the north of the downstream). Extensive glaciers and snow are distributed in the source of the Nyang River and Parlung Zangbo River (Figure 1c). will promote our understanding of vegetation dynamic variations in high-altitude regions and be useful for regional sustainable development.

Study Area
The YZR stretches from west to east through the southern edge of the QTP, with a length of about 2057 km (Figure 1 (a)). It springs from the northwestern part of the YZR basin and is surrounded by lots of mountains, such as Tanggula Shan, Nyainqentanglha Shan, Gangdise Shan, and the Himalayas (Figure 1 (b)). Moreover, there is a significant elevation range, from 133 m to over 7000 m, in the YZR basin. In this study, Nugesha and Nuxia hydrological stations were taken as the outlets of the upstream and midstream, respectively, which divided the YZR basin into three subcatchments. The region between the Chemayungdung Glacier and Nugesha hydrological station is defined as the upstream and is an arid and semi-arid region with annual mean precipitation of less than 300 mm. The region between the Nugesha and Nuxia hydrological stations is defined as the midstream, which has a temperate climate. The region below the Nuxia hydrological station, with an annual mean precipitation of more than 2000 mm, is defined as the downstream characterized by a warm and wet climate. The elevation increases from the southeastern part (the downstream) to northwestern part (the upstream). In addition, there are five tributaries-Dogxung Zangbo River and Nyang Qu River (upstream), Lhasa River and Nyang River (the east of the midstream), and Parlung Tsangpo River (the north of the downstream). Extensive glaciers and snow are distributed in the source of the Nyang River and Parlung Zangbo River (Figure 1 (c)). (a)

GLDAS NOAH Data
Global Land Data Assimilation System (GLDAS) is jointly developed by the Goddard Space Flight Center (GSFC) of the National Aeronautics and Space Administration (NASA) and the National Environmental Prediction Center (NCEP) of the National Oceanic and Atmospheric Administration (NOAA), which combines advanced land surface models and data assimilation techniques to generate global fields of land surface fluxes. In this study, monthly precipitation and surface air temperature data generated by the NOAH land surface model in GLDAS with the spatial resolution of 0.25° [58] were used to supplement the sparse observational network in YZR basin. In addition, we also validated the applicability of GLDAS NOAH data since 2000 in the YZR basin and concluded that GLDAS NOAH data after 2000 can satisfy requirements of analyzing the relationship

GLDAS NOAH Data
Global Land Data Assimilation System (GLDAS) is jointly developed by the Goddard Space Flight Center (GSFC) of the National Aeronautics and Space Administration (NASA) and the National Environmental Prediction Center (NCEP) of the National Oceanic and Atmospheric Administration (NOAA), which combines advanced land surface models and data assimilation techniques to generate global fields of land surface fluxes. In this study, monthly precipitation and surface air temperature data generated by the NOAH land surface model in GLDAS with the spatial resolution of 0.25 • [58] were used to supplement the sparse observational network in YZR basin. In addition, we also validated the applicability of GLDAS NOAH data since 2000 in the YZR basin and concluded that GLDAS NOAH data after 2000 can satisfy requirements of analyzing the relationship between vegetation dynamics and climate change. Related results were shown in Supplementary Section 2.

Vegetation Index Data
The MOD13A3 NDVI product with 1-km spatial resolution and one-month temporal resolution was obtained from the United States Geological Survey (USGS) [59]. This product has been enhanced by reducing external influence factors and inherent non-vegetation influence factors. The remote sensing image processing software ENVI was utilized to transform the HDF-format of this product, extract NDVI values, mosaic the remote sensing images, complete projection conversion and for cutting work. In addition, the NDVI product was further integrated into a 0.25 • spatial resolution in order to be consistent with that of the GLDAS NOAH data in this study. There are different vegetation types in the YZR basin, such as sparse alpine grassland in the upstream region, shrub-grassland in the midstream, and forest in the downstream region [37], and the growing season is defined from April to October, which can efficiently reflect the seasonal cycle of different vegetation types [60]. In this study, the growing season NDVI was applied to indicate the vegetation dynamics in the YZR basin. For the whole basin, the NDVI value of each year was calculated by averaging growing season NDVI values over all pixels.

Linear Regression Model
A linear regression model was applied to detect temporal variation trends of the growing season NDVI [61]. A positive slope (b > 0) means that vegetation is greening while a negative slope (b < 0) means that vegetation is browning. The regression model was as follows: where y is the growing season NDVI; x is the year; b is the regression slope, showing the direction and magnitude of the temporal variation; and a is the regression interception.

Pearson Correlation Analysis
The Pearson correlation coefficient was used to determine the effects of precipitation and surface air temperature on the vegetation cover from 2000 to 2016 in this paper. The method has been widely applied to analyze the correlation between climate factors and NDVI [62,63]. The correlation coefficient (r xy ) is calculated as where n is the number of data pairs; x i and y i are two series of paired data; x and y are the average values of each set of data, respectively; and r xy reflects the consistency between the two variables, ranging from −1 to 1. The absolute value approaching 1 indicates higher consistency.

Partial Correlation Analysis
The partial correlation coefficient can effectively eliminate the quantitative contribution of different climate variables to the variation of growing season NDVI because of the strong correlation between precipitation and surface air temperature [64]. The partial correlation coefficient (r xy,z ) was calculated as follows: r xy,z = r xy − r xz · r yz where r xy,z is the partial correlation coefficient of the variables of x and y, which eliminates the impacts of the variable z. r xy , r xz and r yz are the Pearson correlation coefficients between the variables of x and y, x and z, and y and z, respectively.

Mann-Kendall Nonparametric Test
Sen's slope coupled with the Mann-Kendall significance test has been regarded as an efficient trend analysis method [65], and has been successfully applied in vegetation dynamics [66]. For a data time series of x i = (x 1 , x 2 , . . . , x n ), β, denoting Sen's slope, was calculated as follows: in which 1 < j < i < n. A positive value of β indicates an "upward trend," while a negative value of β indicates a "downward trend." In the Mann-Kendall method, the null hypothesis H 0 states that the data are a sample of independent and identically distributed random variables. The alternative hypothesis H 1 of a two-sided test is that the distribution of x i and x j are not identical for all i, j ≤ n with j i. The test statistic is given as follows: where x i and x j are the sequential data values, n is the length of the time series, and For sample sizes larger than 10, the statistic S is nearly normally distributed, i.e., the statistic is a standard normal variable. Where 18 (8) in which t is the extent of any given tie and denotes the summation over all ties; Var(S) represents the variance of S. When |Z c | > Z (1−α/2) is satisfied where α is the significance level, the null hypothesis H 0 will be rejected, indicating a statistically significant trend.

Contribution Analysis of Human Activities
Residual trend analysis (RESTREND) can differentiate vegetation dynamic variations (denoted by NDVI) caused by human activities from those resulting from climate change [67][68][69]. It is clear that there are two reasonable assumptions involved in RESTREND [62,70,71]. First, vegetation dynamic variations are controlled by human activities and climate change. Second, several climate factors, such as precipitation and surface air temperature, can reflect adequately the influence of climate change in vegetation dynamics. Based on the above assumptions, RESTREND is calculated as follows: where NDVI ob is the observed NDVI values from the MOD13A3 images, reflecting the dual effects of human activities and climate change, and NDVI cl is the predicted NDVI values derived by establishing a linear regression model based on precipitation and surface air temperature. ∆ is the residual value, which can reflect the contribution of human activities on vegetation dynamic variations. Taking an example, if NDVI ob exceeds NDVI cl , residual values are positive, indicating that human activities are beneficial for vegetative growth. On the contrary, negative residual values indicate that human activities are bad for vegetative growth and inhibit the greening trend.

Temporal and Spatial Variations
By averaging growing season NDVI values over all pixels, the linear regression model was used to identify the vegetation temporal variation since 2000. As shown in Figure  Considering the spatial distribution of vegetation, the growing season NDVI values increased gradually from the upstream (northwest) to the downstream (southeast) (Figure 3). Statistical results indicated that the growing season NDVI values of >0.45 accounted for 18.29% area of the YZR basin, most of which appeared in the downstream region, while NDVI values of <0.15 accounted for 25.14% area of the whole basin, most of which appeared in the upstream region. Obviously, there are great geographical differences of growing season NDVI in the YZR basin, which is closely associated with the diverse climate conditions [76,77]. The Indian monsoon is weakened by barrier effect of the mountains, resulting in precipitation and surface air temperature decreasing from the southeast to the northwest of the YZR basin [78], which is consistent with the vegetation spatial distribution pattern (Figure 3), implying that the water and heat characteristics have a considerably important role in vegetation cover. In addition, the vegetation spatial distribution is also determined by vegetation types [79], which was illustrated by the combination of Figures 1c and 3 in the YZR basin. The upstream region is mainly characterized by sparse alpine grassland and the midstream mainly consists of shrub-grassland, while the downstream region is composed of forest indicating the best vegetation cover condition.

Spatial Trend Variations
By combining Sen's slope (Figure 4(a)) and the Mann-Kendall significance test (Figure 4(b)), the growing season NDVI variation trend was classified into five categories: significantly browning, slightly browning, relatively stable, slightly greening and significantly greening (details in Table 1) and its spatial variation trend is shown in Figure 4(c). The proportion of greening regions (including slightly and significantly greening) was over 50% of the YZR basin, which was consistent with the significantly greening tendency at the river basin scale estimated by the linear regression model in Section 3.1.1. It is worth emphasizing that regions with a significantly greening trend accounted for 25.4% area of the basin, mainly concentrated in the Nyang River sub-basin (midstream) and Parlung Tsangpo River sub-basin (downstream). Although Sun et al. used the GIMMS NDVI from 1982 to 2010 to study the vegetation dynamics in the YZR basin and concluded that only the Nyang River sub-basin (midstream) exhibited a significantly greening trend [80], another significantly greening region, the Parlung Tsangpo River sub-basin located in the downstream, was identified in this study. In contrast, the browning regions (including significantly browning and slightly browning) accounted for <25% area of the YZR basin, which were primarily distributed in the urbanized cities of the midstream, implying that human activities played an important role in the vegetation greening process. Regions with a relatively stable NDVI occupied 18.3% area of the basin, which mostly appeared as small and fragment patches in the upstream.

Spatial Trend Variations
By combining Sen's slope (Figure 4(a)) and the Mann-Kendall significance test (Figure 4(b)), the growing season NDVI variation trend was classified into five categories: significantly browning, slightly browning, relatively stable, slightly greening and significantly greening (details in Table 1) and its spatial variation trend is shown in Figure 4(c). The proportion of greening regions (including slightly and significantly greening) was over 50% of the YZR basin, which was consistent with the significantly greening tendency at the river basin scale estimated by the linear regression model in Section 3.1.1. It is worth emphasizing that regions with a significantly greening trend accounted for 25.4% area of the basin, mainly concentrated in the Nyang River sub-basin (midstream) and Parlung Tsangpo River sub-basin (downstream). Although Sun et al. used the GIMMS NDVI from 1982 to 2010 to study the vegetation dynamics in the YZR basin and concluded that only the Nyang River sub-basin (midstream) exhibited a significantly greening trend [80], another significantly greening region, the Parlung Tsangpo River sub-basin located in the downstream, was identified in this study. In contrast, the browning regions (including significantly browning and slightly browning) accounted for <25% area of the YZR basin, which were primarily distributed in the urbanized cities of the midstream, implying that human activities played an important role in the vegetation greening process. Regions with a relatively stable NDVI occupied 18.3% area of the basin, which mostly appeared as small and fragment patches in the upstream.

Spatial Trend Variations
By combining Sen's slope (Figure 4a) and the Mann-Kendall significance test (Figure 4b), the growing season NDVI variation trend was classified into five categories: significantly browning, slightly browning, relatively stable, slightly greening and significantly greening (details in Table 1) and its spatial variation trend is shown in Figure 4c. The proportion of greening regions (including slightly and significantly greening) was over 50% of the YZR basin, which was consistent with the significantly greening tendency at the river basin scale estimated by the linear regression model in Section 3.1.1. It is worth emphasizing that regions with a significantly greening trend accounted for 25.4% area of the basin, mainly concentrated in the Nyang River sub-basin (midstream) and Parlung Tsangpo River sub-basin (downstream). Although Sun et al. used the GIMMS NDVI from 1982 to 2010 to study the vegetation dynamics in the YZR basin and concluded that only the Nyang River sub-basin (midstream) exhibited a significantly greening trend [80], another significantly greening region, the Parlung Tsangpo River sub-basin located in the downstream, was identified in this study. In contrast, the browning regions (including significantly browning and slightly browning) accounted for <25% area of the YZR basin, which were primarily distributed in the urbanized cities of the midstream, implying that human activities played an important role in the vegetation greening process. Regions with a relatively stable NDVI occupied 18.3% area of the basin, which mostly appeared as small and fragment patches in the upstream.

Elevation-Dependent Vegetation Dynamics
The pattern of Elevation Dependent Warming (EDW) is more evident in high-mountain regions, which indicated that these regions experienced more rapid warming [81]. The EDW pattern has great impacts on vegetation variations along the elevation gradient. Therefore, it is of great importance for

Elevation-Dependent Vegetation Dynamics
The pattern of Elevation Dependent Warming (EDW) is more evident in high-mountain regions, which indicated that these regions experienced more rapid warming [81]. The EDW pattern has great impacts on vegetation variations along the elevation gradient. Therefore, it is of great importance for ecological restoration and conservation of the YZR basin with a complex topography (173-7141 m, Figure 1b) to have a deep knowledge of vegetation variation patterns according to elevation. Elevation in this study was divided into intervals of 500 m, and the interval medians were used to represent corresponding intervals. For example, 250 on the x-axis represented the elevation belt of (0, 500), 750 in x-axis represented the elevation belt of (500, 1000), and so on. The elevation dependency of vegetation cover is shown in Figure 5a. The highest growing season NDVI (>0.800) occurred at the elevation belts of <1250 m, corresponding to the downstream region, which also indicated the better vegetation condition in the downstream region, whereas the growing season NDVI decreased with the elevation rising in these belts. Regarding the elevation belts of >1250 m, the growing season NDVI decreased dramatically as the elevation increased, implying an important role of elevation in the spatial distribution of vegetation. Consistent with the decreasing pattern of the growing season NDVI along with the elevated altitude, the vegetation types change from thick forest to sparse vegetation [82]. The largest decrease of the growing season NDVI between adjacent elevation belts occurred from 2750 m to 3250 m, which was attributed to an important transition of vegetation types from the forest in low-altitude regions to alpine grassland in high-altitude regions [83]. The above analyses further emphasize the importance of investigating spatial variation characteristics of the vegetation in the YZR basin.
However, the greening rate of the vegetation exhibited a much more complicated relationship with elevation, which is shown in Figure 5b. The greening rates slightly decreased from elevation belts of 250 m to 1250 m, which was consistent with the NDVI variation within the same elevation belts. Nevertheless, the greening rate exhibited a lifting breakpoint with the highest greening rate of 0.0023/year at the elevation belt of 1750 m and continued to decrease from elevation belts of 2250 m to 3250 m. Similarly, a distinct variation characteristic occurred at the elevation belt from 3000 m to 3500 m, exhibiting the largest declining magnitude between adjacent elevation belts and the only negative greening rate of −0.0037/year, which is consistent with the largest decrease of growing season NDVI at this elevation belt as demonstrated above (Figure 5a). This may be attributed to the urban expansion and agricultural activities owing to the main population centers and crop production region in the elevation belt [84]. Climbing to the transition elevation belt of 3250 m, the greening rate showed a contrary elevation dependency, which increased with the rise of elevation. A more rapid warming tendency is evident in high-altitude regions [85], and the YZR basin has also experienced an evident warming process with a rate of 0.03/year (Supplementary Section 1). The water and energy cycle in this region have become more active and intensive, resulting in glaciers melting and snow cover thawing [86], snow line rising [87], and atmosphere water content enhancement [88], which can provide materials and living space for vegetation recovery and region greening. This may, to some extent, explain the greening rate increasing with elevation lifting above 3500 m. Moreover, although the greening rate decreased with elevation increasing at the lower elevation belts (from 250 m to 3250 m) and increased with elevation increasing at the higher elevation region (above 3500 m), the mean greening rate of the lower elevation belts was still higher than that at the higher elevation belts, which corroborated the greening tendency in the downstream region.

Responses of Vegetation Dynamics to Climate Change
The YZR basin has become warmer and drier according to the data shown in Supplementary Section 1, while precipitation and surface air temperature directly drive the water and energy cycle and further affect the vegetation dynamics. Therefore, it is necessary for the YZR basin, given its complex geographical environment to investigate vegetation responses to these main climate factors.

Temporal Response
Pearson correlation coefficient was applied to quantitatively analyze vegetation responses to annual precipitation and surface air temperature during the period 2000-2016 in the YZR basin, and results are shown in Table 2. The growing season NDVI was significantly positively correlated with surface air temperature (RNDVI-T = 0.472, p < 0.05) but negatively correlated with precipitation (RNDVI-P = −0.339, p > 0.05), which indicated that the surface air temperature was more responsible for the vegetation greening since 2000. Similarly, Zhu et al. indicated that the vegetation greening represented by leaf vegetation index (LAI) in the high-altitude regions, such as the QTP, was more attributed to global warming [4]. Under the background of global warming, the energy balance has been altered greatly, which is more evident in the YZR basin distributed with extensive glaciers and snow cover [89]. Moreover, the snow melting process in this region will produce more shallow moisture and further promote vegetation greening [90][91][92]. In addition, the rising temperature has extended the length of the growing season, promoting vegetation photosynthesis [93], which is also an important reason for the vegetation greening in the YZR. However, the precipitation amount deduction may delay the vegetation phenology [94], which could partly explain the negative correlation between the growing season NDVI and precipitation. In addition, the negative effect of precipitation on the vegetation in this study could also be partly ascribed to the significantly increasing extreme precipitation events in the YZR basin [95,96].

Responses of Vegetation Dynamics to Climate Change
The YZR basin has become warmer and drier according to the data shown in Supplementary Section 1, while precipitation and surface air temperature directly drive the water and energy cycle and further affect the vegetation dynamics. Therefore, it is necessary for the YZR basin, given its complex geographical environment to investigate vegetation responses to these main climate factors.

Temporal Response
Pearson correlation coefficient was applied to quantitatively analyze vegetation responses to annual precipitation and surface air temperature during the period 2000-2016 in the YZR basin, and results are shown in Table 2. The growing season NDVI was significantly positively correlated with surface air temperature (R NDVI-T = 0.472, p < 0.05) but negatively correlated with precipitation (R NDVI-P = −0.339, p > 0.05), which indicated that the surface air temperature was more responsible for the vegetation greening since 2000. Similarly, Zhu et al. indicated that the vegetation greening represented by leaf vegetation index (LAI) in the high-altitude regions, such as the QTP, was more attributed to global warming [4]. Under the background of global warming, the energy balance has been altered greatly, which is more evident in the YZR basin distributed with extensive glaciers and snow cover [89]. Moreover, the snow melting process in this region will produce more shallow moisture and further promote vegetation greening [90][91][92]. In addition, the rising temperature has extended the length of the growing season, promoting vegetation photosynthesis [93], which is also an important reason for the vegetation greening in the YZR. However, the precipitation amount deduction may delay the vegetation phenology [94], which could partly explain the negative correlation between the growing season NDVI and precipitation. In addition, the negative effect of precipitation on the vegetation in this study could also be partly ascribed to the significantly increasing extreme precipitation events in the YZR basin [95,96]. As shown in Table 2, precipitation and surface air temperature in the growing season showed a strong relationship, with R P-T = −0.523, so it is necessary to eliminate the autocorrelation effect between precipitation and surface air temperature to examine the quantitative contribution of these two climate factors to the variation of vegetation. The partial correlation coefficient (PR) was used in this study and PR NDVI-P and PR NDVI-T were calculated yearly according to Equation (3). As shown in Figure 6a, PR NDVI-P was between −0.253 and 0.582 with an average value of 0.162, while the partial correlation coefficient between NDVI and surface air temperature (PR NDVI-T ) was 0.143-0.724, with an average value of 0.525. The proportion of PR NDVI-T > 0.600 accounted for 52.9%, while partial correlation coefficients between NDVI and precipitation were rather lower and negative values occurred in the year of 2000, 2006, and 2016. This reemphasized the more attributable role of the surface air temperature for improving and enhancing the vegetation coverage in the YZR basin.
Additionally, there was a strongly negative correlation between PR NDVI-P and PR NDVI-T (Figure 6b As shown in Table 2, precipitation and surface air temperature in the growing season showed a strong relationship, with RP-T = −0.523, so it is necessary to eliminate the autocorrelation effect between precipitation and surface air temperature to examine the quantitative contribution of these two climate factors to the variation of vegetation. The partial correlation coefficient (PR) was used in this study and PRNDVI-P and PRNDVI-T were calculated yearly according to Equation (3). As shown in Figure  6 Additionally, there was a strongly negative correlation between PRNDVI-P and PRNDVI-T ( Figure  6

Spatial Response
Regarding spatial correlation, great heterogeneity of the correlation between the growing season NDVI and precipitation (Figure 7(a)) and surface air temperature (Figure 7(b)) were illustrated. Regions with a positive correlation coefficient between the growing season NDVI and surface air temperature were located in the Nyang River sub-basin (midstream) and Parlung Tsangpo River subbasin (downstream), where the vegetation experienced a significantly greening tendency ( Figure  4(c)). As mentioned above, the sources of these sub-basins are distributed with extensive glaciers and snow cover, and more shallow soil moisture from warming-induced glacier and snow melting are beneficial for vegetation growth. Li et al. also concluded that soil moisture at the depth of 0−10 cm from the GLDAS NOAH showed an increasing trend in the YZR basin since the 1970s [97]. Regions with a negative correlation coefficient between the growing season NDVI and surface air temperature were relatively sporadically distributed in the upstream.
Similar to the contrary temporal variation patterns of PRNDVI-P and PRNDVI-T (Figure 6), the spatial distribution patterns of correlation coefficients between the growing season NDVI and precipitation were also opposite to those between the growing season NDVI and surface air temperature. That is,

Spatial Response
Regarding spatial correlation, great heterogeneity of the correlation between the growing season NDVI and precipitation (Figure 7a) and surface air temperature (Figure 7b) were illustrated. Regions with a positive correlation coefficient between the growing season NDVI and surface air temperature were located in the Nyang River sub-basin (midstream) and Parlung Tsangpo River sub-basin (downstream), where the vegetation experienced a significantly greening tendency (Figure 4c). As mentioned above, the sources of these sub-basins are distributed with extensive glaciers and snow cover, and more shallow soil moisture from warming-induced glacier and snow melting are beneficial for vegetation growth. Li et al. also concluded that soil moisture at the depth of 0−10 cm from the GLDAS NOAH showed an increasing trend in the YZR basin since the 1970s [97]. Regions with a negative correlation coefficient between the growing season NDVI and surface air temperature were relatively sporadically distributed in the upstream.
Similar to the contrary temporal variation patterns of PR NDVI-P and PR NDVI-T (Figure 6), the spatial distribution patterns of correlation coefficients between the growing season NDVI and precipitation were also opposite to those between the growing season NDVI and surface air temperature. That is, regions with a positive correlation coefficient between the growing season NDVI and precipitation were mainly concentrated in the arid and semi-arid upstream. Annual precipitation of less than 300 mm in the upstream limits vegetation growth because moisture masses from the Indian Ocean are blocked by many mountains [37]. Therefore, precipitation from atmospheric moisture is necessary to support regional ecosystem stability in the arid and semi-arid upstream region [98], which is responsible for the positive correlation in the upstream region. regions with a positive correlation coefficient between the growing season NDVI and precipitation were mainly concentrated in the arid and semi-arid upstream. Annual precipitation of less than 300 mm in the upstream limits vegetation growth because moisture masses from the Indian Ocean are blocked by many mountains [37]. Therefore, precipitation from atmospheric moisture is necessary to support regional ecosystem stability in the arid and semi-arid upstream region [98], which is responsible for the positive correlation in the upstream region. Figure 7. The spatial distribution of the correlation coefficient between NDVI and precipitation (a) and NDVI and temperature (b). The white point represents a significantly positive correlation at the 0.05 significance level.

Responses of Vegetation Dynamics to Human Activities
In addition to climate change, human activities, mainly characterized by urbanization and agriculture activities in the YZR basin, also exert significant impacts on vegetation spatiotemporal variations. Therefore, investigating the spatial responses of vegetation in the high-altitude regions to human activities is of great importance.
According to Equation (

Responses of Vegetation Dynamics to Human Activities
In addition to climate change, human activities, mainly characterized by urbanization and agriculture activities in the YZR basin, also exert significant impacts on vegetation spatiotemporal variations. Therefore, investigating the spatial responses of vegetation in the high-altitude regions to human activities is of great importance.
According to Equation ( For the whole YZR basin, the annual growing season NDVI residuals showed a significantly decreasing trend with a rate of −0.003/year at the 95% confidence level (Figure 8b), mainly ascribed to the transition from positive residual values to negative residual values. This indicates an overall negative impact of human activities on vegetation, which should be considered seriously in future agricultural production and socioeconomic development.  [100]. These findings both indicated that human activities represented by agricultural and urban developments played an important role in vegetation dynamic variations. For the whole YZR basin, the annual growing season NDVI residuals showed a significantly decreasing trend with a rate of −0.003/year at the 95% confidence level (Figure 8(b)), mainly ascribed to the transition from positive residual values to negative residual values. This indicates an overall negative impact of human activities on vegetation, which should be considered seriously in future agricultural production and socioeconomic development.     [100]. These findings both indicated that human activities represented by agricultural and urban developments played an important role in vegetation dynamic variations. For the whole YZR basin, the annual growing season NDVI residuals showed a significantly decreasing trend with a rate of −0.003/year at the 95% confidence level (Figure 8(b)), mainly ascribed to the transition from positive residual values to negative residual values. This indicates an overall negative impact of human activities on vegetation, which should be considered seriously in future agricultural production and socioeconomic development.

Discussion
In addition to precipitation and surface air temperature, the spatial variation trend of the soil moisture in the YZR basin was also analyzed, because soil moisture constrains vegetation's evapotranspiration and photosynthesis processes and plays an important role in vegetation dynamic variations. The results showed that the areas with a significantly positive trend were mainly distributed in the Nyang River and Parlung Zangbo River basins, denoted as white points in Figure 10, which was consistent with the areas exhibiting the positive relationship between surface air temperature and vegetation (Figure 7b). Significant glacier shrinkage has been found in these sub-basins, which brings abundant soil moisture and promotes vegetation growth [101]. Therefore, it can be deduced that glacier and snow melting resulted in more soil moisture from 2000 to 2016, and this process further promoted vegetation growth.

Discussion
In addition to precipitation and surface air temperature, the spatial variation trend of the soil moisture in the YZR basin was also analyzed, because soil moisture constrains vegetation's evapotranspiration and photosynthesis processes and plays an important role in vegetation dynamic variations. The results showed that the areas with a significantly positive trend were mainly distributed in the Nyang River and Parlung Zangbo River basins, denoted as white points in Figure  10, which was consistent with the areas exhibiting the positive relationship between surface air temperature and vegetation (Figure 7(b)). Significant glacier shrinkage has been found in these subbasins, which brings abundant soil moisture and promotes vegetation growth [101]. Therefore, it can be deduced that glacier and snow melting resulted in more soil moisture from 2000 to 2016, and this process further promoted vegetation growth. Precipitation, as the main and direct input of soil moisture, may exert a hysteresis effect on vegetation dynamic variations, and GLDAS precipitation and soil moisture (0-100 cm) were therefore used to explore the effect of soil moisture on vegetation at different time lags (Figure 11). The highest correlation between NDVI and precipitation was observed when time lag = 1 (R = 0.898), indicating that precipitation at a one-month leading time exhibited the most important influence on vegetation growth. However, the highest correlation between NDVI and soil moisture was observed when time lag = 0 (R = 0.866). Compared with soil moisture, precipitation required a one-month time lag to reach the best status to improve vegetation growth at the river basin scale. It can be inferred that precipitation might be contained underground and transformed into soil moisture after one month to be beneficial for vegetation growth in the YZR basin. Precipitation, as the main and direct input of soil moisture, may exert a hysteresis effect on vegetation dynamic variations, and GLDAS precipitation and soil moisture (0-100 cm) were therefore used to explore the effect of soil moisture on vegetation at different time lags (Figure 11). The highest correlation between NDVI and precipitation was observed when time lag = 1 (R = 0.898), indicating that precipitation at a one-month leading time exhibited the most important influence on vegetation growth. However, the highest correlation between NDVI and soil moisture was observed when time lag = 0 (R = 0.866). Compared with soil moisture, precipitation required a one-month time lag to reach the best status to improve vegetation growth at the river basin scale. It can be inferred that precipitation might be contained underground and transformed into soil moisture after one month to be beneficial for vegetation growth in the YZR basin.

Discussion
In addition to precipitation and surface air temperature, the spatial variation trend of the soil moisture in the YZR basin was also analyzed, because soil moisture constrains vegetation's evapotranspiration and photosynthesis processes and plays an important role in vegetation dynamic variations. The results showed that the areas with a significantly positive trend were mainly distributed in the Nyang River and Parlung Zangbo River basins, denoted as white points in Figure  10, which was consistent with the areas exhibiting the positive relationship between surface air temperature and vegetation (Figure 7(b)). Significant glacier shrinkage has been found in these subbasins, which brings abundant soil moisture and promotes vegetation growth [101]. Therefore, it can be deduced that glacier and snow melting resulted in more soil moisture from 2000 to 2016, and this process further promoted vegetation growth. Precipitation, as the main and direct input of soil moisture, may exert a hysteresis effect on vegetation dynamic variations, and GLDAS precipitation and soil moisture (0-100 cm) were therefore used to explore the effect of soil moisture on vegetation at different time lags (Figure 11). The highest correlation between NDVI and precipitation was observed when time lag = 1 (R = 0.898), indicating that precipitation at a one-month leading time exhibited the most important influence on vegetation growth. However, the highest correlation between NDVI and soil moisture was observed when time lag = 0 (R = 0.866). Compared with soil moisture, precipitation required a one-month time lag to reach the best status to improve vegetation growth at the river basin scale. It can be inferred that precipitation might be contained underground and transformed into soil moisture after one month to be beneficial for vegetation growth in the YZR basin. In this study, we investigated vegetation dynamic variations and their associated responses to environmental changes in the YZR basin from 2000 to 2016, based on the MODIS NDVI and GLDAS data. Results showed that the vegetation had a significantly increasing trend since 2000, and significantly, the greening trend mainly concentrated in the Parlung Zangbo River basin and Nyang River basin, indicating the importance of glacier and snow melting processes in the YZR basin. However, there are some limitations and uncertainties in the current study. First, although precipitation and surface air temperature are the dominant climate factors of vegetation variations in the YZR basin, other relevant factors should be taken into consideration for explicit investigation on the interaction mechanism between climate change and vegetation dynamics. Zhao et al. [102] showed that ENSO controlled vegetation dynamic variations at the global scale, while geographical differences existed, i.e., vegetation growth was mainly determined by the temperature at northern high-latitudes, by water in arid and semi-arid regions, and by radiation in the Amazon region. Nemina et al. [103] considered that radiation explained 27% of the vegetation greenness. Second, the response to climate change of the vegetation cover in high-altitude regions is more sensitive, which may be related to the elevation-dependent gradients of ecosystem properties, such as vegetation types, species components and soil structures [88], further showing the importance of the elevation. Yao et al. indicated that the precipitation wetting trend was amplified with elevation, which was attributed to the accelerating warming-driven water circulation [104]. Piao et al. pointed out that the elevation played an important role in controlling the spatial patterns of the vegetation green-up date [105]. Moreover, one significant problem associated with investigating elevation-dependent gradients is that more intensive human activities in high-altitude areas will cause some troubles in understanding the interactive responses between vegetation dynamics and climate change. Hence, the effects of elevation-dependent gradients on distributing vegetation greening and environmental factors should be a great focus. Third, climate change often involves nonlinear, nonstationary complex processes with periodic oscillations [106], while vegetation dynamic variations are often controlled by climate change, showing similar nonlinear, nonstationary processes [107]. However, the non-linear responses between vegetation cover and climate change in the YZR basin are still unclear. Finally, the residual analysis method based on multiple linear regression may also undervalue the contribution of climate change on vegetation dynamic variations [108], and it is necessary to use other advanced methods to explicitly separate the contributions of climate change and human activities.

Conclusions
In this study, it was revealed that the YZR basin had experienced a significant vegetation greening process with an increasing rate of 0.001/year (Z c = 2.31, p < 0.05) since 2000, and regions with significant greening were mainly concentrated in the Nyang River sub-basin (midstream) and Parlung Tsangpo River sub-basin (downstream). Analysis on the elevation-dependent vegetation showed that the vegetation cover of the low-altitude regions was better than that of the high-altitude regions, while the greening rate exhibited a much more complicated relationship with the elevation.
The surface air temperature was shown to play a more important role in vegetation greening, and significantly positive correlations between the growing season NDVI and surface air temperature were found in the Nyang River and Parlung Zangbo sub-basins. There are extensive glaciers and snow cover in these regions that have experienced significantly increasing trends of soil moisture since 2000. It can be deduced that soil moisture from glaciers and snow melting played an important role in vegetation growth. Although basin-wide non-significant negative correlations were found between precipitation and growing season NDVI, positive influences of precipitation on vegetation greening occurred in the arid and semi-arid upstream region. Additionally, the one-month time lag of precipitation to reach the optimal status to improve vegetation growth at the basin scale indicated that precipitation may be contained underground and transformed into soil moisture, thereby playing a role in vegetation growth.
In addition to precipitation and surface air temperature, the associated responses of other climate factors, such as radiation and snow water equivalent and the cooling effect caused by increased vegetation cover, should be paid more attention in the future.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-4292/11/20/2421/s1, Figure S1: Temporal variations of the annual precipitation and surface air temperature during the growing season in the YZR basin since 2000, Figure S2: GLDAS NOAH and observed precipitation (a) and air temperature (b) variations, Figure S3: The correlation coefficient between GLDAS NOAH and in situ precipitation (a) and surface air temperature (b), Table S1: Statistical indicators of precipitation (Pre) and surface air temperature (Tem) between GLDAS NOAH and in situ data.