The impact of drought on vegetation conditions within the Damqu River Basin, Yangtze River Source Region, China

Abstract

Drought and vegetation conditions within the Damqu River Basin, part of the Yangtze River Source Region (YRSR), are assessed here using the standardized precipitation index (SPI), the standardized precipitation evapotranspiration index (SPEI), the normalized difference vegetation index (NDVI), and the leaf area index (LAI). We utilized Sen’s method, least squares regression method, linear regression and Pearson’s correlation analysis to study variations in drought and vegetation indices and the drought effect on vegetation between 1988 and 2015. Results reveal that droughts occurred at a 25% frequency over this period; SPI and SPEI analyses show that 1994, 1999, 2005, and 2010 were change points and that the basin was characterized by varying drought and humidity trends. Subsequent to 2010, both SPI and SPEI decreased within the basin, while 1995, 2000, 2004, and 2010 were change points for NDVI and LAI while the watershed exhibited variable trends in vegetation reduction and increase. The NDVI-annual values of 63.36% regions and the LAI-summer values of 68.39% areas within the basin were decreased during 1988–2015 and 2000–2015, respectively. Subsequent to 2010, both NDVI and LAI decreased within the basin and significant positive correlations at inter-annual and inter-summer time scales were seen in both drought and vegetation indices; drought has exerted a lag effect on vegetation as shown by significant positive correlations between annual SPI/SPEI values and following year NDVI/LAI values.

About the Authors:

Zhilong Zhao

Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

Affiliations Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China

Yili Zhang

Roles Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing

Affiliations Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China, CAS Centre for Excellence in Tibetan Plateau Earth Sciences, Beijing, China

Linshan Liu

Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Validation, Writing – review & editing

* E-mail: [email protected]

Affiliation: Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China

ORCID logo http://orcid.org/0000-0002-0194-2214

Zengzeng Hu

Roles Validation, Writing – review & editing

Affiliation: College of Urban Economics and Public Administration, Capital University of Economics and Business, Beijing, China

Introduction

High-latitude regions are amongst the most sensitive to climate change globally [1–3]. As the “third pole” of the Earth, the water tower of Asia, and the national ecological security shelter zone of China [4–8], the Qinghai-Tibet Plateau faces the dual threats of climate change and human development [8–9]. The Three River Source Region (TRSR) within the hinterland of the Qinghai-Tibet Plateau is one of the largest areas of natural wetland at highest altitudes nationally and is often referred to as the “Chinese Water Tower”. This region is also critical for national ecological security and water conservation even though it comprises a vulnerable and sensitive ecological environment. The Damqu River Basin is located in the southwest of the Three River Source Nature Reserve and is the southern source of the Yangtze River.

China experiences frequent and severe droughts [10–12] that are especially marked within the TRSR [13]. Climate data for the period between 1971 and 2010 reveals that the TRSR can nevertheless be characterized by significant regional differences; although the climate of northern and eastern regions has been predominantly warm and humid since the 1990s, southern and western areas have remained mainly warm and dry [14]. It is also the case that global warming scenarios predict that droughts are likely to become more frequent and severe in the future [15]; these events negatively impact the growth of grasslands, reducing their productivity [16], and are a major reason for land degradation [17–20]. Droughts also represent a considerable threat to the sustainable development of animal husbandry economic systems, especially on grasslands [16]. This is relevant because one previous field study carried out within the Damqu River Basin between 2015 and 2016 showed that southwestern region of the TRSR is susceptible to aridification [21]. The American Meteorological Society classified drought-related phenomena into four types, meteorological, agricultural, hydrological, and socioeconomical [22]. Researchers often apply a variety of indices in this context to characterize complex drought phenomena, the most common of which include the standardized precipitation index (SPI) [23], the standardized precipitation evapotranspiration index (SPEI) [10,16,24–28], the Palmer drought severity index (PDSI) [29,30], the integrated drought condition index (IDCI) [31], the standardized vegetation index (SVI) [32], the optimized meteorological drought index (OMDI), and the optimized vegetation drought index (OVDI) [33]. The first two of these, the SPI and SPEI, have the advantage of incorporating multiple time scales (i.e., one month, three months, six months, 12 months, 24 months, and 36 months [24]) and are therefore more suitable than their counterparts for assessing the sensitivity of vegetation communities to water deficits [34]. In addition, the SPEI also evaluates the positive influence of precipitation and the negative influence of evapotranspiration on drought intensity [16,24] while at the same time remaining sensitive to the negative influence of temperature in the same way as the SPI and PDSI [16].

Our current understanding of how ecosystems respond to droughts remains incomplete and requires further research [35]. Variations in vegetation dynamics are often used to characterize ecosystem changes and are one important proxy that can be used to gain an in-depth understanding of global system fluctuations, especially in the context of climate change [35]. We therefore used the normalized difference vegetation index (NDVI) and leaf area index (LAI) in this study to characterize vegetation conditions within the Damqu River Basin in addition to the SPI and the SPEI to characterize drought conditions. We explored the responses of vegetation dynamics to drought conditions within the Damqu River Basin on the basis of these indices, initially assessing the inter-annual variability of events over a timescale of 12 months. We also assessed the inter-summer variability of droughts over a three-month timescale encompassing the wettest season of the year.

Materials and methods

Study area

The Damqu River is 331 km long and has a basin perimeter enclosed within 92.14° - 94.62° E and 32.40° - 33.96° N. This river basin is located at elevations between 4,503 m and 5,947 m and encompasses a total area of about 1.67 × 104 km2. This region has a continental climate, with an average annual temperature of -1.05°C and an annual precipitation of 444.53 mm; records from the China Meteorological Data Center (CMDC) collected between 1998 and 2015 show that 60% of total precipitation occurs in the summer, in June, July, and August. These observations also show that the rate of annual mean temperature increase has been 0.07°C per year (significance (α) = 0.001), and that values were enhanced significantly between 2003 and 2015. Similarly, the annual precipitation increase rate has been 2.32 mm per year (α = 0.05); 1994, 1999, 2005, and 2010 were times when trends changed, and precipitation tended to either decrease or increase between these points [21]. We have shown in previous work that the area of wetlands decreased within the Damqu River Basin between 1988 and 2015 [21], and assume that drought is the main reason for this reduction. In order to test this hypothesis, we used the same timescale (between 1988 and 2015) to assess drought and vegetation condition.

A total of 13 land-cover types can be defined within the Damqu River Basin including Kobresia swampy meadow, steppe meadow, alpine desert, K. pygmaea wet meadow, Carex-Kobresia swamp, alpine steppe, flooded wetland, glacier, river, lacustrine pond, lake, bare land, and glacial lake (Fig 1). Land cover data for this region in 2015 [21] show that the area of Kobresia swampy meadow accounted for the highest proportion (25.62%), while glacial lakes just cover 0.01% of the basin. The total area of the six main land cover types (i.e., Kobresia swampy meadow, steppe meadow, alpine desert, K. pygmaea wet meadow, Carex-Kobresia swamp, and alpine steppe) accounted for 95.47% of the basin area in 2015.

[Figure omitted. See PDF.]

Fig 1. Map to show the location and land cover distribution of the Damqu River Basin watershed.

https://doi.org/10.1371/journal.pone.0202966.g001

Data sources

Remote sensing (RS) data.

The RS data utilized in this study includes a Global Inventory Monitoring and Modeling Systems (GIMMS3g) NDVI dataset, a MOD13A2 NDVI dataset, and a MOD15A2 LAI dataset (Table 1). We used two of these datasets (GIMMS3g and MOD13A2) to compute NDVI changes and to assess whether, or not, they exhibit the same characteristics during the same period.

1. The GIMMS3g NDVI dataset used in this study was downloaded from https://ecocast.arc.nasa.gov/. Data were subjected to maximum value synthesis using the maximum value composite (MVC) approach in order to mitigate interference from clouds, the atmosphere, and solar altitude angles [36]. Annual NDVI was defined using average values for the 12 months in each calendar year.

2. Variation in NDVI across the growing season (i.e., June, July, and August, the wettest season of the year) between 2000 and 2014 was extracted from the MOD13A2 data product (https://ladsweb.nascom.nasa.gov/data/search.html). These data were also subjected to MVC analysis to minimize interference from clouds, the atmosphere, and solar altitude angles [37]. In this case, annual growing season NDVI was defined using average values for the period between May 25th and August 28th of each calendar year.

3. Variations in LAI throughout the growing season (i.e., June, July, and August) between 2000 and 2015 were calculated using MOD15A2 data (ftp://ftp.glcf.umd.edu/glcf/GLASS/LAI/MODIS/). Interference due to cloud pollution and the atmosphere were removed and missing data were supplemented [38]; annual growing season LAI was defined using average values for the period between May 25th and August 28th of each calendar year.

[Figure omitted. See PDF.]

Table 1. The main characteristics of RS products used in this study.

https://doi.org/10.1371/journal.pone.0202966.t001

Meteorological data.

The meteorological data used in this study were acquired from the CMDC (http://data.cma.cn/site/index.html) and include monthly mean air temperature and precipitation between 1988 and 2015. All data downloaded from the CMDC were evaluated using strict quality controls and checks, tested for erroneous entries, and subjected to homogeneity correction by the National Meteorological Information Center of China. We selected seven weather stations (S1 Table) in the vicinity of the Damqu River Basin for meteorological data collection as there are none within this region (Fig 1). Average monthly precipitation values were used to calculate the SPI, while average monthly temperature and precipitation values were used to calculate the SPEI.

Methods

SPI and SPEI.

The SPI was developed by Mckee et al. [39] and is one of the key indicators that is widely used to characterize variations in meteorological and hydrological droughts. We used the calculation software spi_sl_6.exe (http://drought.unl.edu/MonitoringTools/DownloadableSPIProgram.aspx) provided by the National Drought Mitigation Center to calculate the SPI.

The SPEI was developed by Vicente-Serrano et al. [24] and is also commonly used to characterize drought variation. This index was calculated using the software provided by Vicente–Serrano et al. [24] (http://digital.csic.es/handle/10261/10002).

Statistical analyses.

We used the ratio of a region involved droughts (Pi) to assess the frequency of drought [40,41]:(1)

Where n is the number of the years in a region that involved droughts based on SPI or SPEI (SPI or SPEI ≤ −0.50) [16,27], N is the number of the years during 1988 to 2015, and i is the code for a region.

We used the nonparametric Mann-Kendall test for detecting the significance of the increasing or decreasing trend and the nonparametric Sen’s method for evaluating the slope of a linear trend. The Mann-Kendall test requires at least 4 values and calculation of the confidence intervals for the Sen’s slope estimate requires at least 10 values in a time series of values of meteorological elements [42,43].

We used the least squares method in order to detect trends in each index and to perform piecewise linear fitting of the SPI, the SPEI, the NDVI, and the LAI. This approach has been applied in previous work to identify overall trends and to compute the breakpoints between periods with significantly different tendencies [44]. We applied the statistical procedure and operation code outlined by Tomé and Miranda [44] in this study using the software ENVI/IDL to identify trend change points in drought and vegetation time series. These points divided the drought and vegetation time series in this study into some sub-periods, which can help find the stage characteristics in different sub-periods. We then performed a correlation analysis once these points had been obtained to assess the statistical significance of SPI, SPEI, NDVI, and LAI values.

We used the slope of as the index to detect the trend in vegetation dynamics for every pixel by using the least squares method, which can synthetically reflect the spatiotemporal change characteristics of vegetation coverage [45]. The slope is calculated as follows [45]:(2)where Slope is the trend of vegetation coverage variation, n is the number of the years that RS products used in this study have crossed, i is the order of year from 1 to n, and NDVIi or LAIi is the value of NDVI or LAI in the ith year. When Slope>0, the NDVI or LAI shows an increasing trend, and when Slope<0, the NDVI or LAI shows a decreasing trend [45].

We calculated Pearson’s correlation coefficients between drought and vegetation indices at three and 12 month time scales and assessed their statistical significance at 0.01, 0.05, and 0.1 confidence levels for different months and years.

Results

SPI variation

Inter-annual variation in the SPI between 1988 and 2015 (Fig 2A) shows that droughts occurred in seven years according to the standard in Table 2 within the Damqu River Basin, at a 25% frequency. Data show that mild droughts occurred in 1990, 1995, 1997, and 2006, a moderate drought occurred in 1992, and severe droughts occurred in 1994 and 2015. Based on the Sen’s estimate, the values of SPI-annual showed a rise with a linear rate of 0.03/a during 1988–2015, and it passed the significance level α = 0.05 through Mann-Kendall test (Fig 2A).

[Figure omitted. See PDF.]

Fig 2.

Variations in the SPI obtained from average monthly precipitation values and SPEI obtained from average monthly temperature and precipitation values within the Damqu River Basin between 1988 and 2015 (a) SPI-annual (b) SPEI-annual (c) SPI-summer (d) SPEI-summer.

https://doi.org/10.1371/journal.pone.0202966.g002

[Figure omitted. See PDF.]

Table 2. SPI and SPEI categories [16,27].

https://doi.org/10.1371/journal.pone.0202966.t002

The least squares method was used to obtain trend change points in SPI time series, and then we detected the stage characteristics of the SPI between these points. Trend point analysis results show that changes of SPI-annual occurred in 1994, 1999, 2005, and 2010 (Fig 3A); SPI decreased between 1988 and 1994 (Pearson’s correlation coefficients (r) = -0.601, significance (p) = 0.077), indicating an increased drought trend during this period. In contrast, SPI increased between 1994 and 1999 (r = 0.896, p = 0.008), indicating an increased humidification trend during this period, while this index decreased again between 1999 and 2005 (r = -0.346, p = 0.201) indicating an enhanced drought trend. SPI also increased between 2005 and 2010 (r = 0.219, p = 0.338) indicating enhanced humidification during this period, while values of this index decreased from 2010 (r = -0.422, p = 0.202), indicating the more risk of drought.

[Figure omitted. See PDF.]

Fig 3.

Point analysis of change trends and stage characteristics in the SPI and SPEI within the Damqu River Basin between 1988 and 2015 (a) SPI-annual (b) SPEI-annual (c) SPI-summer (d) SPEI-summer.

https://doi.org/10.1371/journal.pone.0202966.g003

Inter-summer (between June and August) variation in the SPI over the time period of this study (Fig 2C) shows that droughts occurred in seven years at a 25% frequency. Mild droughts occurred in 1990, 1992, 1996, and 2006, moderate droughts occurred in 1997 and 2005, and severe droughts occurred in 1994. Based on the Sen’s estimate, the values of SPI-summer showed an increase with a linear rate of 0.02/a during 1988–2015, and it passed the significance level α = 0.1 through Mann-Kendall test (Fig 2C).

Point analysis shows that trend changes of SPI-summer occurred in 1994 and 2009 (Fig 3C), while inter-summer SPI conformed to a downward trend between 1988 and 1994 (r = -0.447, p = 0.158), indicating the more risk of drought during this period. Inter-summer SPI increased between 1994 and 2009 (r = 0.698, p = 0.001) indicating a trend towards higher humidification, while inter-summer SPI decreased from 2009 onwards (r = -0.445, p = 0.158), indicating an increased risk of drought during this period.

SPEI variation

Inter-annual variation in the SPEI between 1988 and 2015 (Fig 2B) shows that droughts occurred in six years according to the standard in Table 2 within the Damqu River Basin, at a 21.42% frequency. Data show that mild droughts occurred in 1992, 1995, 2006, and 2012, while moderate droughts occurred in 1994 and 2015. Based on the Sen’s estimate, the values of SPEI-annual showed a slight rise during 1988–2015, and it did not pass the significance level α = 0.1 through Mann-Kendall test (Fig 2B).

The least squares method was used to obtain trend change points in SPEI time series, and then we detected the stage characteristics of the SPEI between these points. Point analysis shows that trend changes of SPEI-annual occurred in 1994, 2000, 2005, and 2009 (Fig 3B); indeed, inter-annual SPEI followed a downward trend between 1988 and 1994 (r = -0.504, p = 0.124), indicating an enhanced risk of drought during this period, while an increase in inter-annual SPEI between 1994 and 2000 is indicative of an increased humidification trend (r = 0.821, p = 0.012). Inter-annual SPEI again decreased (r = -0.452, p = 0.154) between 2000 and 2005, indicating the more risk of drought, while inter-annual SPEI increased again (r = 0.336, p = 0.290) between 2005 and 2009, evidencing an enhanced humidification trend. Inter-annual values for this index then decreased from 2009 onwards (r = -0.396, p = 0.189), indicating the more risk of drought.

Inter-summer (between June and August) variation in the SPEI over the time period of this study (Fig 2D) reveals that droughts occurred in seven years within the basin, at a 25% frequency. Data show that mild droughts occurred in 1990, 1996, 1997, and 2013, while moderate droughts occurred in 1994, 2006, and 2015. Based on the Sen’s estimate, the values of SPEI-summer showed a slight increase during 1988–2015, and it did not pass the significance level α = 0.1 through Mann-Kendall test (Fig 2D).

Point analysis shows that trend changes of SPEI-summer occurred in 1995 and 2008 (Fig 3D); inter-summer SPEI values declined between 1988 to 1995 (r = -0.385, p = 0.173), indicating an enhanced drought trend throughout this period, while inter-summer SPEI values increased between 1995 and 2008 (r = 0.539, p = 0.023), indicating enhanced humidification. Inter-summer SPEI values decreased again from 2008 onwards (r = -0.534, p = 0.086), further evidencing the more risk of drought.

NDVI variation

Variations in the NDVI are considered separately here on the basis of GIMMS and MODIS data.

NDVI based on GIMMS data.

We explore NDVI variation within the Damqu River Basin between 1988 and 2015 from three perspectives in this section, changes in average values, variation slope of different values and proportional variations within each group with different values. Average values in this case mainly denote overall NDVI trends within the watershed throughout the study period; thus, annual NDVI data were uniformly graded to explore tendencies in variation within high-value and low-value regions in each year and to analyze differences in vegetation growth within the Damqu River Basin. Variation slope of annual NDVI is another aspect to detect spatial differences of vegetation index in the study area during 1988–2015.

Data show that the average annual NDVI value was 0.23 and it showed a slight decrease (r = -0.122, p = 0.268) between 1988 and 2015 (Fig 4).

[Figure omitted. See PDF.]

Fig 4. Variation in average annual NDVI values obtained from GIMMS3g and point analysis of trend changes within the Damqu River Basin between 1988 and 2015.

https://doi.org/10.1371/journal.pone.0202966.g004

The least squares method was used to obtain trend change points in NDVI time series, and then we detected the stage characteristics of the NDVI between these points. The results of point analysis show that trend changes occurred in 1995, 2000, 2005, and 2010 (Fig 4); the average watershed value of NDVI decreased between 1988 and 1995 (r = -0.5040, p = 0.2090), indicating that vegetation was degraded over this period. In contrast, the average watershed NDVI value increased by 0.02 between 1995 and 2000 (r = 0.7811, p = 0.0666), but decreased again between 2000 to 2005 (r = -0.8281, p = 0.0418), demonstrating a trend towards degraded vegetation within the basin. Watershed NDVI increased again between 2005 and 2010 (r = 0.4140, p = 0.4144), but decreased from 2010 onwards (r = -0.6761, p = 0.1404), indicating further vegetation degradation.

The condition of the Damqu River Basin as characterized by GIMMS data is reflected in a maximum NDVI value of 0.40 and a minimum value of 0.05 for the period between 1988 and 2015. A spatial map of average NDVI values for all regions within the basin is shown in Fig 5A, and another spatial map of variation slope of annual NDVI during 1988–2015 is exhibited in Fig 5B. In this period, the NDVI values of 63.36% regions within the basin was decreased, while the NDVI values of the 36.64% regions increased and these regions concentrated in the northwest and central part of the basin. Regional values on Fig 5A can be divided into four groups based on differences between minimum and maximum NDVI values. We then subsequently assessed whether, or not, the proportional changes seen in each group exhibit the same temporal variation characteristics compared to patterns in average NDVI value variation within the Damqu River Basin between 1988 and 2015.

[Figure omitted. See PDF.]

Fig 5.

Average annual NDVI values (a) and NDVI-annual variation slope (b) within the Damqu River Basin between 1988 and 2015.

https://doi.org/10.1371/journal.pone.0202966.g005

The data presented in Fig 6 show that regional areas with NDVI values between 0.1 and 0.2 increased between 1988 and 1995 from 16.81% to 30.17%. This area then subsequently decreased between 1995 and 2000, a proportional decline of 21.12%, before increasing again between 2000 and 2005, a proportional expansion of 23.71%. This area then decreased to 12.50% between 2005 and 2010 before expanding again from 2010 onwards, reaching a proportion of 23.71% by 2015. Trends in regional areas with NDVI values between 0.2 and 0.3 exhibited the opposite behavior between 1988 and 2015 compared to those with values between 0.1 and 0.2, while those with values between 0.3 and 0.4 fluctuated and decreased over this period, an overall decrease in area proportion from 14.66% to 4.74%. Regions with NDVI values between 0.1 and 0.2 obviously expanded subsequent to 2010, while those with values between 0.2 and 0.3 followed a reversed trend. Proportional changes in regions with NDVI values between 0.3 and 0.4 exhibited the same temporal characteristics of variation as average values within the Damqu River Basin between 1988 and 2015. It is therefore clear that these changes influenced variation in average NDVI values across the whole basin.

[Figure omitted. See PDF.]

Fig 6. Proportional variation in annual NDVI values at different levels within the Damqu River Basin between 1988 and 2015.

https://doi.org/10.1371/journal.pone.0202966.g006

NDVI based on MODIS data.

We explore NDVI variation within the Damqu River Basin between 1988 and 2015 from the same three perspectives as discussed for GIMMS data, changes in average values, variation slope of different values and proportional variations within each group with different values. Average values are reported here in the same way as for GIMMS data.

Data show that between 2000 and 2014, the average summer NDVI value was 0.40 and it showed a slight increase (r = 0.295, p = 0.143) (Fig 7).

[Figure omitted. See PDF.]

Fig 7. Variations in average summer NDVI values obtained from MOD13A2 and point analysis of trend changes within the Damqu River Basin between 2000 and 2014.

https://doi.org/10.1371/journal.pone.0202966.g007

Point analysis shows a change between 2004 and 2009 (Fig 7). Average watershed NDVI values between 2000 and 2004 tended to decrease (r = -0.4160, p = 0.4860), indicating vegetation degradation over this period, and this change characteristic is consistent with the annual NDVI value obtained from GIMMS3g in the same period. Average values increased between 2004 and 2009 (r = 0.4884, p = 0.3257) to 0.43, and this variation characteristic is consistent with the annual NDVI value between 2005 and 2009. Then the average values decrease again between 2009 and 2014 (r = -0.6023, p = 0.2058), further evidencing vegetation degradation within the basin and this change characteristic is consistent with the annual NDVI value during 2010–2014.

Summer NDVI values for the Damqu River Basin based on MODIS data reveal a maximum value of 0.77 and a minimum value of -0.18 for the period between 2000 and 2014. A spatial map of average summer NDVI values for all regions within the basin is presented in Fig 8A, and another spatial map of change slope of summer NDVI during 2000–2014 is showed in Fig 8B. In this period, the NDVI values of 25.46% regions within the basin was decreased, while the NDVI values of the 74.54% regions increased. Then we divided the values for all regions into five groups (Fig 8A) based on differences between minimum and maximum summer NDVI values and assessed whether, or not, proportional changes within each exhibit the same temporal variations compared to fluctuations in average records between 2000 and 2014. It is also noteworthy that the NDVI categories shown in Fig 8A differ from those in Fig 5A because of the use of different data products; these two figures can therefore be used to analyze regional NDVI differences within the Damqu River Basin.

[Figure omitted. See PDF.]

Fig 8.

Average summer NDVI values (a) and NDVI-summer variation slope (b) within the Damqu River Basin between 2000 and 2014.

https://doi.org/10.1371/journal.pone.0202966.g008

The data presented in Fig 9 show that the regional area characterized by summer NDVI values between 0.2 and 0.4 increased between 2000 and 2004, from 46.08% to 47.27%. Subsequently, this area decreased between 2004 and 2009 to 28.56%, and then increased again from 2009 onwards, reaching 44.07% in 2014. In contrast, summer NDVI values between 0.4 and 0.6 between 2000 and 2014 followed the opposite trend, while regional area with values between 0.6 and 0.8 fluctuated to some extent. This area fluctuated and increased between 2000 and 2009 as the proportion of such regions within the watershed area increased from 3.97% to 6.62%. The proportion of this area decreased from 2009 onwards to 2.72% by 2014; data show that regions with NDVI values between 0.2 and 0.4 have expanded over time, while those with values between 0.4 and 0.6 have followed the opposite trend. Proportional changes in regions with NDVI values between 0.4 and 0.6 have exhibited the same temporal variations as average values within the Damqu River Basin between 2000 and 2014; it is clear that these changes have influenced variations in average NDVI values across the whole basin.

[Figure omitted. See PDF.]

Fig 9. Proportional variation in summer NDVI values at different levels within the Damqu River Basin between 2000 and 2014.

https://doi.org/10.1371/journal.pone.0202966.g009

Variation in the LAI

Trends in the LAI within the Damqu River Basin between 2000 and 2015 are explored here from three perspectives, changes in average values, variation slope of different values and proportional variations within each group with different values. Average values in this case mainly denote overall trends within the watershed over the study period, while LAI data were uniformly graded in order to explore variations in both high- and low-value regions in each year and to analyze differences in vegetation growth and intensity within the Damqu River Basin. Variation slope of LAI is another aspect to detect spatial differences of vegetation index in the study area during 2000–2015.

Data show that the average summer LAI value was 6.40 and it showed a slight increase (r = 0.101, p = 0.355) between 2000 and 2015 (Fig 10).

[Figure omitted. See PDF.]

Fig 10. Variation in average summer LAI values obtained from MOD15A2 and point analysis of trend changes within the Damqu River Basin between 2000 and 2015.

https://doi.org/10.1371/journal.pone.0202966.g010

The least squares method was used to obtain trend change points in LAI time series, and then we detected the stage characteristics of the LAI between these points. Point analysis reveals that trend changes occurred in 2004 and 2010 (Fig 10). Average watershed LAI values between 2000 and 2004 tended to decrease (r = -0.7220, p = 0.1684), indicating vegetation degradation over this period. Average watershed LAI values increased, however, between 2004 and 2010 (r = 0.5642, p = 0.1871) to 0.93, but decreased again between 2010 and 2015 (r = -0.6221, p = 0.1872); this latter trend is indicative of degraded vegetation within the basin.

Damqu River Basin MODIS data results reveal a maximum summer LAI value of 27 and a minimum value of zero between 2000 and 2015. A spatial map of average summer values for all regions across the basin is presented in Fig 11A, and another spatial map of variation slope of summer LAI during 2000–2015 is showed in Fig 11B. In this period, the LAI values of 68.39% regions within the basin was decreased, while the LAI values of the 31.61% regions increased and these regions concentrated in the northwest and south part of the basin. These data on Fig 11A show that the proportional area of regions with LAI values greater than 15 accounts for 0.15% of the whole basin. We therefore divided all LAI values for all regions into five groups (Fig 11A), based on differences between minimum and maximum summer values and their proportional areas. We then assessed whether, or not, proportional changes for each group exhibit the same temporal characteristics of variation when compared to average summer values across the Damqu River Basin between 2000 and 2015.

[Figure omitted. See PDF.]

Fig 11.

Average summer LAI values (a) and LAI-summer variation slope (b) within the Damqu River Basin between 2000 and 2015.

https://doi.org/10.1371/journal.pone.0202966.g011

The data presented in Fig 12 show that the regional area characterized by summer LAI values up to five increased between 2000 and 2004, but decreased between 2004 and 2010; this area proportion decreased overall from 38.89% to 26.61%. This area then increased again after 2010, reaching 41.27% by 2015. In contrast, trends in the proportion of LAI values between five and ten between 2000 and 2015 followed the opposite pattern compared to their counterparts up to five, while the area characterized by values more than ten fluctuated slightly. The proportional area of these regions within the watershed then decreased from 8.15% to 4.70% between 2000 and 2015; from 2010 onwards, regions characterized by LAI values up to five tended to expand, while those characterized by values between five and ten expanded relatively obviously and those between five and ten contracted. Proportion changes in regional areas characterized by LAI values less than five have exhibited the same temporal characteristics as average values within the Damqu River Basin between 2000 and 2015; it is clear that these changes have influenced variations in average LAI values across the whole basin.

[Figure omitted. See PDF.]

Fig 12. Proportional variation in summer LAI values at different levels within the Damqu River Basin between 2000 and 2015.

https://doi.org/10.1371/journal.pone.0202966.g012

Correlation between SPI and NDVI/LAI

Observations show that vegetation coverage within the Damqu River Basin increased in humid years and decreased in arid ones. Data show that 2009 vegetation correspond with peak drought indices (Fig 9), while the trough seen in 1995 corresponds to the same pattern in drought indices (Fig 6).

We performed a correlation analysis on SPI values for each year versus NDVI and LAI values for the same year. Results, however, revealed no significance at p = 0.1; the same result was also recovered for SPI values for each year versus LAI values for the same year. At the same time, analyses show that SPI values are negatively correlated with both NDVI and LAI values; a higher and more positive SPI value equates to more humid meteorological conditions that are increasingly beneficial for vegetation growth, while no negative correlations between SPI and NDVI/LAI were recovered. We also performed a correlation analysis to evaluate SPI values for each year versus NDVI/LAI values for the following year (Table 3).

[Figure omitted. See PDF.]

Table 3. Pearson’s correlation coefficients for vegetation indices and SPI values within the Damqu River Basin.

https://doi.org/10.1371/journal.pone.0202966.t003

The results presented in Table 3 reveal significant positive correlations between SPI and NDVI/LAI over a one year lag temporal scale. At the same time, however, responses in vegetation indices (NDVI/LAI) to SPI exhibited time lag phenomena; both the NDVI and LAI exhibited significant positive responses to the SPI during summer, demonstrating that droughts have a more pronounced impact on values of these indices during the summer than on annual NDVI. The fact that the correlation between NDVI and SPI was significantly higher than between LAI and SPI shows that the former is more sensitive to variations in drought indices.

Correlation between SPEI and NDVI/LAI

We performed additional correlation analyses to assess SPEI values in each year versus NDVI and LAI values for the same year. Results show that SPEI values for each year and NDVI values for the same year were not significantly correlated (p = 0.1); this was also the case for annual SPEI values and LAI values for the same year although values of the former were negatively correlated with both the NDVI and LAI. In this case, a higher and more positive SPEI value is indicative of more humid meteorological conditions, favoring plant growth. Data also reveal an absence of negative correlations between SPEI and NDVI/LAI; we therefore performed a correlation analysis to compare the SPEI values for each year versus NDVI/LAI values for the next (Table 4).

[Figure omitted. See PDF.]

Table 4. Pearson’s correlation coefficients for vegetation indices and SPEI within the Damqu River Basin.

https://doi.org/10.1371/journal.pone.0202966.t004

Significant positive correlations exist between SPEI and NDVI/LAI at one year lag temporal scales (Table 4). At the same time, responses in vegetation indices (NDVI/LAI) to SPEI also exhibit time lag phenomena; both NDVI and LAI exhibited significant positive responses to the SPEI during the summer which shows that drought exerts a more pronounced impact on summer NDVI and LAI values than annual values of the former. Data show a significantly higher level of correlation between NDVI and SPEI than between LAI and SPEI; this suggests that NDVI is more sensitive to variations in drought indices.

Discussion

Droughts are amongst most damaging of natural disasters from socioeconomic, environmental, and human perspectives [46]. Data show that the spatial extent of droughts in northern and eastern Asia increased between 1950 and 2000 [46], and that a dipole-type configuration with northerly droughts and southerly flooding was seen in eastern Asia between 1901 and 2014 [47]. The frequency of droughts has increased over the 21st century [36] in most parts of China, while significant dry trends have been mainly seen in the southwest, north, northwest, and central parts of the country. The frequency of severe droughts in China remained extremely high in the 1990s and 2000s and drought frequency had increased in Tibetan Plateau since 1980s [10], while northern and eastern parts of the TRSR have experienced increased humidification since the 1990s and from the 21st century onwards, respectively. In contrast, other parts of TRSR have experienced trends towards continuous droughts since the early 1980s; this has especially been the case in southern and western regions where wetness index linear trend rates have reached -0.8% per year [14]. As discussed, the Damqu River Basin is located within the southwestern TRSR; because of alternating dry-wet conditions throughout growing seasons between 2001 and 2010, most areas of the TRSR have been subject to humidification while the southern part of Zaduo County (containing the Damqu River Basin) has experienced a tendency towards droughts [48]. In this study, we also found the Damqu River Basin have experienced drought for six or seven times since the 1990s and from the 21st century onwards. One consequence of drought is that the water status of this region has changed, with possible consequences for grassland productivity; declining grassland productivity constitutes a threat to the sustainable development of animal husbandry, the major economic activity within the Damqu River Basin. This watershed is also the southern source of the Yangtze River within China and comprises vast wetland areas; subsequent reductions in water contents have also negatively impacted the Damqu River Bain environment.

Vegetation cover in China has increased since the 1980s [49–51]. And Tibetan Plateau is also greening under the positive impact of climate change [51]. Data show that Tibetan Plateau NDVI values gradually increased between 1982 and 2012 up until the end of the 1990s and then decreased slightly in subsequent years [52] as a result of grassland variations [53]. The maximum NDVI recorded on this plateau increased in the summer (between June and August) by 0.4% over the ten years between 1982 and 2013 [54], while the spatial distribution of this index exhibits remarkable longitude zonality and increases continuously from west to east on the Tibetan Plateau [55]. Areas of grassland and forests decreased between 1990 and 2010 in the central high Himalayas [56]. Similarly, the area of degraded meadows in southern Qinghai Province increased 20.36% during 1976–2015 [57]. After the implementation of ecological protection measures [58,59] and the construction of the TRSR, more than 60% of the grassland area showed a significant improvement [60]. In light of the impacts of regional climate variation, vegetation dynamics also experience significant changes at the regional level; western and southern central regions have experienced vegetation degradation, for example, while northern and northwestern areas have seen increases [61]. Regions characterized by degraded vegetation are mainly distributed in Qumarleb and Zaduo counties within the eastern and southern parts of the TRSR [62–64], including the Damqu River Basin. In this study, we also found the NDVI-annual values of 63.36% regions within the Damqu River Basin was decreased during 1988–2015. In China, the growing-season (April–October) LAI (LAIGS) showed an increase (average trend of 0.0070/a, ranging from 0.0035/a to 0.0127/a) during 1982–2009 [49]. The average summer LAI value also showed a slight increase with a linear rate of 0.0086/a between 2000 and 2015 within the Damqu River Basin, even though it did not pass the significance level α = 0.1. Based on the satellite datasets of GIMMS, GLOBMAP and GLASS, the change trend of LAIGS in most regions of the Damqu River Basin was between -0.01 and 0.01 during the period 1982–2009 [49]. In this study, we also found the variation slope of LAI-summer values of 61.03% regions within the Damqu River Basin was between -0.01 and 0.01 during 2000–2015 (Fig 11B).

We utilized two different datasets (i.e., GIMMS3g and MOD13A2) in this analysis to compute NDVI changes and generated different results [65]. Data show that while NDVI values for the Damqu River Basin fluctuated between 0.21 and 0.24 (GIMMS3g), watershed values ranged between 0.38 and 0.43 (MOD13A2). Changes in NDVI nevertheless exhibited the same temporal characteristics between 2000 and 2015 irrespective of dataset.

The results of this study corroborate previous related research in the identification of lag effects in vegetation responses to drought [16,36]. It is notable that the drought indices used in this study were calculated from precipitation and temperature data; it remains an open question whether, or not, lag effects within the Damqu River Basin are the result of temperature or precipitation and further research in this area will be required. Limitations in both the density and spatial distribution of meteorological data also raise uncertainties about the accuracy of SPI and SPEI results for the Damqu River Basin.

Conclusions

We explored the dynamics of drought and vegetation cover across the Damqu River Basin in this study and assessed the impacts of the former on the latter between 1998 and 2015.

The results of this study show that both SPI and SPEI can be used as reliable indicators to characterize hydrological drought conditions within the Damqu River Basin. These two indicators exhibit the same temporal characteristics over the study period; between 1988 and 2015, droughts occurred at a frequency of 25%. Trend change points were seen in 1994, 1999, 2005, and 2010; indicators between these points exhibited both downward and upward trends when the watershed was subject to a higher frequency of droughts or increased humidification, respectively. Subsequent to 2010, both parameters decreased, indicating the more risk of drought within this basin.

Results show that both the NDVI and LAI can be used as indicators to characterize vegetation conditions within the Damqu River Basin; both these indices displayed the same temporal characteristics over the period of this study. The NDVI-annual values of 63.36% regions within the basin was decreased during 1988–2015. Similarly, the LAI-summer values of 68.39% regions within the basin was decreased between 2000 and 2015. Change points in trends were seen in 1995, 2000, 2004, and 2010; indicators at these times either decreased or increased when the watershed was subject to vegetation degradation or enhanced growth, respectively. The fact that both parameters decreased after 2010 is also indicative of degraded vegetation.

The data presented in this study reveal that vegetation coverage within the Damqu River Basin grew better in humid years while the opposite was the case in arid years. Drought and vegetation indices are both characterized by significant positive correlations at different time scales across the Damqu River Basin. It is noteworthy that a time lag effect in the responses of both NDVI and LAI vegetation indices was seen with respect to the SPI and SPEI drought indices.

Supporting information

[Figure omitted. See PDF.]

S1 Table. Locations of weather stations in the vicinity of the Damqu River Basin.

https://doi.org/10.1371/journal.pone.0202966.s001

(DOCX)

S2 Table. Detailed values of drought indices within the Damqu River Basin during 1988–2015.

https://doi.org/10.1371/journal.pone.0202966.s002

(DOCX)

S3 Table. Detailed values of vegetation indices within the Damqu River Basin during 1988–2015.

https://doi.org/10.1371/journal.pone.0202966.s003

(DOCX)

Acknowledgments

The authors would like to thank Dr. Zhaofeng Wang and Dr. Lanhui Li of Institute of Geographic Sciences and Natural Resources Research of CAS, and Dr. Haiwei Zhao of Institute of Botany of CAS for their thoughtful suggestions on the manuscript.

Citation: Zhao Z, Zhang Y, Liu L, Hu Z (2018) The impact of drought on vegetation conditions within the Damqu River Basin, Yangtze River Source Region, China. PLoS ONE 13(8): e0202966. https://doi.org/10.1371/journal.pone.0202966

References

1. Pan BT, Li JJ. Qinghai-Tibetan Plateau: a driver and amplifier of the global climatic change. Journal of Lanzhou University (Natural Sciences). 1996; 32(1): 108–115. (in Chinese)

2. Yao T. TPE international program: a program for coping with major future environmental challenges of The Third Pole region. Progress in Geography. 2014; 33(7): 884–892. (in Chinese)

3. Smol JP. Lakes and Climate Change—a Paleoecological Perspective. Dordrecht: Springer; 2014.

4. Yanai MH, Li CF, Song ZS. Seasonal heating of the Tibetan Plateau and its effects on the evolution of the asian summer monsoon. Journal of the Meteorological Society of Japan. 1992; 70: 319–351.

5. Immerzeel WW, van Beek LPH, Bierkens MFP. Climate Change Will Affect the Asian Water Towers. Science. 2010; 328: 1382–1385. https://doi.org/10.1126/science.1183188 pmid:20538947

6. Sun HL, Zheng D, Yao TD, Zhang YL. Protection and construction of the National Ecological Security Shelter Zone on Tibetan Plateau. Acta Geographica Sinica. 2012; 67(1): 3–12. (in Chinese)

7. Yao T, Thompson L, Yang W, Yu W, Gao Y, Guo X, et al. Different glacier status with atmospheric circulations in Tibetan Plateau and surroundings. Nature Climate Change. 2012a; 2: 663–667. https://doi.org/10.1038/nclimate1580

8. Yao T, Thompson LG, Mosbrugger V, Zhang F, Ma Y, Luo T, et al. Third Pole Environment (TPE). Environmental Development. 2012b; 3: 52–64. https://doi.org/10.1016/j.envdev.2012.04.002

9. IPCC WG II. The Contribution to the IPCC's Fifth Assessment Report. Cambridge: Cambridge University Press; 2014.

10. Liu X, Wang S, Zhou Y, Wang F, Li W, Liu W. Regionalization and Spatiotemporal Variation of Drought in China Based on Standardized Precipitation Evapotranspiration Index (1961–2013). Advances in Meteorology. 2015; 2015:1–18. https://doi.org/10.1155/2015/950262

11. Chen T, Xia G, Liu T, Chen W, Chi D. Assessment of Drought Impact on Main Cereal Crops Using a Standardized Precipitation Evapotranspiration Index in Liaoning Province, China. Sustainability. 2016; 8: 1069. https://doi.org/10.3390/su8101069

12. Zeng X, Zhao N, Sun H, Ye L, Zhai J. Changes and Relationships of Climatic and Hydrological Droughts in the Jialing River Basin, China. PLoS ONE. 2015; 10(11): e0141648. https://doi.org/10.1371/journal.pone.0141648 pmid:26544070

13. Ren X, Dong Z, Hu G, Zhang D, Li Q. A GIS-Based Assessment of Vulnerability to Aeolian Desertification in the Source Areas of the Yangtze and Yellow Rivers. Remote Sensing. 2016; 8(8): 626. https://doi.org/10.3390/rs8080626

14. Xu WX, Gu S, Su WJ, Jiang S, Xiao RX, Xiao JS, et al. Spatial pattern and its variations of aridity/humidity during 1971–2010 in Three-River Source Region on the Qinghai-Tibet Plateau. Arid Land Geography. 2012; 35(1): 46–55. (in Chinese)

15. Dai AG. Drought under global warming: A review. Wiley Interdisciplinary Reviews Climate Change. 2011; 2: 45–65.

16. Tong S, Bao Y, Te R, Ma Q, Si H, Lusi A. Analysis of Drought Characteristics in Xilingol Grassland of Northern China Based on SPEI and Its Impact on Vegetation. Mathematical Problems in Engineering. 2017; 2017:1–11. https://doi.org/10.1155/2017/5209173

17. Nicholson SE, Tucker CJ, Ba MB. Desertification, drought and surface vegetation: An example from the west African Sahel. Bulletin of the American Meteorological Society. 1998; 79: 815–829.

18. Pickup G. Desertification and climate change—The Australian perspective. Climate Research. 1998; 11: 51–63.

19. Vicente-Serrano S, Cabello D, Tomás-Burguera M, Martin-Hernandez N, Begueria S, Azorin-Molina C, et al. Drought Variability and Land Degradation in Semiarid Regions: Assessment Using Remote Sensing Data and Drought Indices (1982–2011). Remote Sensing. 2015; 7(4): 4391–4423. https://doi.org/10.3390/rs70404391

20. Liu P, Hao L, Pan C. Combined effects of climate and land management on watershed vegetation dynamics in an arid environment. Science of the Total Environment. 2017; 589: 73–88. https://doi.org/10.1016/j.scitotenv.2017.02.210 pmid:28264774

21. Zhao ZL. Dynamic of wetland in Damqu River Basin in the source region of the Yangtze River and the influencing factors. Beijing: Univ. Chinese Academy of Sciences; 2017. (in Chinese)

22. Mishra AK, Singh VP. A review of drought concepts. Journal of Hydrology. 2010; 391(1–2): 202–216.

23. Mondol AH, Ara I, Das SC. Meteorological Drought Index Mapping in Bangladesh Using Standardized Precipitation Index during 1981–2010. Advances in Meteorology. 2017; 2017: 1–17. https://doi.org/10.1155/2017/4642060

24. Vicente-Serrano S M, Beguería S, López-Moreno JI. A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index. Journal of Climate. 2010; 23: 1696–1718.

25. Ivits E, Horion S, Fensholt R, Cherlet M. Global Ecosystem Response Types Derived from the Standardized Precipitation Evapotranspiration Index and FPAR3g Series. Remote Sensing. 2014a; 6(5): 4266–4288. https://doi.org/10.3390/rs6054266

26. Törnros T, Menzel L. Addressing drought conditions under current and future climates in the Jordan River region. Hydrology and Earth System Sciences. 2014; 18: 305–318. https://doi.org/10.5194/hess-18-305-2014

27. Li B, Zhou W, Zhao Y, Ju Q, Yu Z, Liang Z, et al. Using the SPEI to Assess Recent Climate Change in the Yarlung Zangbo River Basin, South Tibet. Water. 2015a; 7(10): 5474–5486. https://doi.org/10.3390/w7105474

28. Li Z, Zhou T, Zhao X, Huang K, Wu H, Du L. Diverse spatiotemporal responses in vegetation growth to droughts in China. Environmental Earth Sciences. 2016; 75: 55. https://doi.org/10.1007/s12665-015-4781-0

29. Palmer WC. Meteorological Drought. U.S. Weather Bureau, Washington, D. C.; 1965.

30. Dai AG, Kevin ET, Qian TT. A Global Dataset of Palmer Drought Severity Index for 1870–2002: Relationship with Soil Moisture and Effects of Surface Warming. Journal of Hydrometeorology. 2004; 5: 1117–1130.

31. Meng L, Dong T, Zhang W. Drought monitoring using an Integrated Drought Condition Index (IDCI) derived from multi-sensor remote sensing data. Natural Hazards. 2016; 80: 1135–1152. https://doi.org/10.1007/s11069-015-2014-1

32. Shah SU, Iqbal J. Spatial–temporal variations of vegetation and drought severity across Tharparkar, Pakistan, using remote sensing-derived indices. Journal of Applied Remote Sensing. 2016; 10(3): 036005. https://doi.org/10.1117/1.JRS.10.036005

33. Hao C, Zhang J, Yao F. Combination of multi-sensor remote sensing data for drought monitoring over Southwest China. International Journal of Applied Earth Observation and Geoinformation. 2015; 35: 270–283.

34. Vicente-Serrano SM, Beguería S, Lorenzo-Lacruz J, Camarero JJ, López-Moreno JI, Azorin-Molina C, et al. Performance of drought índices for ecological, agricultural and hydrological applications. Earth Interactions. 2012; 16: 1–27.

35. Ivits E, Horion S, Fensholt R, Cherlet M. Drought footprint on European ecosystems between 1999 and 2010 assessed by remotely sensed vegetation phenology and productivity. Glob Chang Biol. 2014b; 20(2): 581–593. https://doi.org/10.1111/gcb.12393

36. Li Z, Zhou T, Zhao X, Huang K, Gao S, Wu H, et al. Assessments of Drought Impacts on Vegetation in China with the Optimal Time Scales of the Climatic Drought Index. International Journal of Environmental Research and Public Health. 2015b; 12: 7615–7634. https://doi.org/10.3390/ijerph120707615

37. Deng LC, Liu SD, Zhao XY. Study on the change in land cover of Yangtze River Delta based on MOD13A2 data. China Science Paper. 2015; 10(15): 1822–1827. (in Chinese)

38. Myneni RB, Hoffman S, Knyazikhin Y, Privette JL, Glassy J, Tian Y, et al. Global products of vegetation leaf area and fraction absorbed PAR from year one of MODIS data. Remote Sensing of Environment. 2002; 83: 214–231.

39. McKee TB, Doesken NJ, Kleist J. The relation ship of drought frequency and duration to time scales. California: Eighth Conference on Applied Climatology. 1993; 179–184.

40. He JN, Li YG, Li X, Huang JC. Temporal and spatial characteristics of droughts over Yunnan Province during 1961–2012. Mountain Research. 2016; 34(1): 19–27. (in Chinese)

41. Huang WH, Yang XG, Li MS, Zhang XY, Wang MT, Dai SW, et al. Evolution characteristics of seasonal drought in the south of China during the past 58 years based on standardized precipitation index. Transactions of the Chinese Society of Agricultural Engineering. 2010; 26(7): 50–59.

42. Kahya E, Kalayci S. Trend analysis of streamflow in Turkey. Journal of Hydrology. 2004; 289(1): 128–144.

43. Tabari H, Talaee P H. Temporal variability of precipitation over Iran: 1966–2005. Journal of Hydrology. 2011; 396(3): 313–320.

44. Tomé AR, Miranda PMA. Piecewise linear fitting and trend changing points of climate parameters. Geophysical Research Letters. 2004; 31: L02207. https://doi.org/10.1029/2003G019100

45. Liu XF, Zhang JS, Zhu XF, Pan YZ, Liu YX, Zhang DH, et al. Spatiotemporal changes in vegetation coverage and its driving factors in the Three-River Headwaters Region during 2000–2011. Journal of Geographical Sciences. 2014; 24(2): 288–302.

46. Sheffield J, Wood EF. Global Trends and Variability in Soil Moisture and Drought Characteristics, 1950–2000, from Observation-Driven Simulations of the Terrestrial Hydrologic Cycle. Journal of Climate. 2008; 21(3): 432–458.

47. Preethi B, Mujumdar M, Kripalani RH, Prabhu A, Krishnan R. Recent trends and tele-connections among South and East Asian summer monsoons in a warming environment. Climate Dynamics. 2016; 48(7–8): 1–17.

48. Wang M, Zhou CP, Wu L, Xu XL, Ou YH. Wet-drought pattern and its relationship with vegetation change in the Qinghai-Tibetan Plateau during 2001–2010. Arid Land Geography. 2013; 36(1): 49–56. https://doi.org/10.13826/j.cnki.cn65-1103/x.2013.01.001 (in Chinese)

49. Piao SL, Yin GD, Tan JG, Cheng L, Huang MT, Li Y, et al. Detection and attribution of vegetation greening trend in China over the last 30 years. Global Change Biology. 2015; 21: 1601–1609. https://doi.org/10.1111/gcb.12795 pmid:25369401

50. Li Y, Piao SL, Li Laurent Z X, Chen AP, Wang XH, Ciais P, et al. Divergent hydrological response to large-scale afforestation and vegetation greening in China. Science Advances. 2018; 4: eaar4182. pmid:29750196

51. Zhu ZC, Piao SL, Myneni R B, Huang MT, Zeng ZZ, Canadell J G, et al. Greening of the Earth and its drivers. Nature Climate Change. 2016; 6:791–795.

52. Pang G, Wang X, Yang M. Using the NDVI to identify variations in, and responses of, vegetation to climate change on the Tibetan Plateau from 1982 to 2012. Quaternary International. 2017; 444: 87–96.

53. Wang G, Han L. The Vegetation NDVI Variation Trend in Qinghai-Tibet Plateau and Its Response to Climate Change. International Conference on Remote Sensing, Environment and Transportation Engineering, IEEE. 2012; 1–4.

54. Sun J, Qin X. Precipitation and temperature regulate the seasonal changes of NDVI across the Tibetan Plateau. Environmental Earth Sciences. 2016; 75(4): 1–9.

55. Wang W, Feng Q, Yu H, Liang T, Guo N. Spatio-temporal change of vegetation on Tibetan Plateau based on AVHRR-NDVI data. Geoscience and Remote Sensing Symposium, IEEE. 2016; 1374–1377.

56. Zhao ZL, Wu X, Zhang YL, Gao JG. Assessment of changes in the value of ecosystem services in the Koshi River Basin, Central High Himalayas based on land cover changes and the CA-Markov Model. Journal of Resources and Ecology. 2017; 8(1): 67–76.

57. Li XL, Gao J, Zhang J, Sun HQ. Natural and anthropogenic influences on the spatiotemporal change of degraded meadows in southern Qinghai Province, West China: 1976–2015. Applied Geography. 2018; 97: 176–183.

58. Lu XG, Wang RF. Study on wetland biodiversity in China. Chinese Geographical Science. 1996; 6(1): 15–23.

59. Hua YY, Cui BS, He WJ, Cai YZ. Identifying potential restoration areas of freshwater wetlands in a river delta. Ecological Indicators. 2016; 71: 438–448.

60. Cai HY, Yang XH, Xu XL. Human-induced grassland degradation/restoration in the central Tibetan Plateau: The effects of ecological protection and restoration projects. Ecological Engineering. 2015; 83: 112–119.

61. Yu XJ, Yan Q, Liu ZJ, Xi LH, Wang Y. Quantitative estimation and dynamic changes of vegetation fractional coverage in the Sanjiangyuan Region. Resources and Environment in the Yangtze Basin. 2013; 22(1): 66–74. (in Chinese)

62. Zhao J, Jiang QG, Li WD, Li YH. Study on evolutionary regionalization of ecological environment in source regions of the three rivers based on changes of NDVI. Global Geology. 2008; 27(4): 427–431. (in Chinese)

63. Liu XF, Zhang JS, Zhu XF, Pan YZ, Liu YX, Zhang DH, et al. Spatiotemporal changes in vegetation coverage and its driving factors in the Three-River Headwaters Region during 2000–2011. Journal of Geographical Sciences. 2014; 24(2): 288–302. https://doi.org/10.1007/s11442-014-1088-0

64. Wu ZF, Li F, Zhang LB, Zhang JP, Du JJ. Research on grassland degradation of Three-River Headwaters region based on reference vegetation coverage. Journal of Natural Disasters. 2014; 23(2): 94–102. (in Chinese)

65. Chen YL, Long BJ, Pan XB, Zhong SQ, Mo WH. Differences between MODIS NDVI and AVHRR NDVI in monitoring grasslands change. Journal of Remote Sensing. 2011; 15(4): 831–845.