The Impacts of Drought Changes on Alpine Vegetation during the Growing Season over the Tibetan Plateau in 1982–2018

Abstract

The Tibetan Plateau (TP) is a climate-sensitive and ecologically fragile area. Studying drought and its effects on vegetation over the TP is of great significance for ecological conservation. However, there were large uncertainties in previous studies on the drought characteristics and their impacts on alpine vegetation in this region. This study explored the drought changes and their impacts on alpine vegetation during the growing season over the TP in 1982–2018. The results showed that the TP has experienced a wetting trend in most regions of the TP. Correspondingly, the vegetation has become greener in most areas. The wetting and drying trend in the growing season changed around 1995. Before 1995, the TP experienced an overall drying trend with a spatial pattern of a drying trend in the northern regions and a wetting trend in the southern regions, while it showed an overall wetting trend after 1995, with a reversed spatial pattern to that before 1995. After 1995, wetting and drying trends affected the vegetation in 61% of the TP. However, before 1995, the NDVI presented an increasing trend in most areas of the TP under a drying trend. Therefore, a drying trend was not the primary factor affecting vegetation growth in this period. Instead, changes in the cryosphere induced by warming could be the main factor. In addition, the distribution of vegetation across the TP was primarily influenced by drought intensity, which had the greatest impact on sparse vegetation, followed by meadow and grassland. This study enhances our understanding of the impact of drought changes on alpine vegetation on the TP.

1. Introduction

Drought is one of the most serious natural disasters faced by humanity [1]. The frequency, duration, and intensity of droughts have tended to increase over the 21st century in the context of climate warming [2], which will not only bring water scarcity and significant losses to agricultural production, but also reduce vegetation productivity and increase plant mortality [3]. These effects can ultimately lead to intensified desertification and ecosystem degradation [4,5,6]. Vegetation is an essential component of terrestrial ecosystems, serving as an indicator of global change by connecting the soil, atmosphere, and water [7]. For the past few years, the impacts of drought on vegetation have drawn more and more attention [8].

The Tibetan Plateau (TP), which is considered the center of the Third Pole, has the highest elevation in the world [9]. Its unique terrain and climate conditions, with minimal human interference, make it particularly susceptible to the impacts of global change [10], and it is considered a premonitory region for global climate change [11]. Over the last few decades, the TP exhibited a significant warming trend at a rate of 0.34 °C/decade, far exceeding the Northern Hemisphere warming rate over the same time period [12]. The most prevalent vegetation type over the TP is alpine grasslands, which cover approximately two-thirds of the TP area [13,14,15]. These grasslands have a low resistance to climatic extremes [3]. Ecosystems over the TP are particularly susceptible to climate change and can be readily devastated by extreme droughts [11]. Consequently, it is significant that we study the effects of drought changes on vegetation growth on the TP, as it is an ecologically critical area in Asia and the world [14].

Studies have shown that changes in vegetation were significantly and positively correlated with drought changes in nearly 54% of global vegetated lands, as well as in approximately 71% of the total area of China [16,17]. Drought affected vegetation mainly through the induction of plant water stress, which resulted in negative impacts on vegetation growth and metabolism, ultimately leading to a slowing of vegetation growth and even the mortality of vegetation [18,19]. Over the TP, some researchers have indicated that the intensity, duration, and frequency of droughts in the TP have shown an upward trend over the past few years [20,21]. This was primarily due to an increase in evapotranspiration induced by upward air temperature (Ta), which finally led to significant reductions in soil moisture and runoff [22,23,24]. Correspondingly, a significant decrease in the vegetation cover and biomass of alpine meadows [25], as well as a decrease in vegetation NPP [26], were found. However, other studies have shown a slowing drought trend in the TP [27,28,29]. The dominant factor for this deceleration was the increasing trend of precipitation (P) along with the climate warming [30,31,32]. At the same time, the NDVI and NPP of vegetation increased based on satellite data and numerical models [33,34]. Furthermore, the responses of vegetation to drought changes presented various instances of spatial heterogeneity in the TP, which varies with the different vegetation types in different regions [14,35,36,37,38,39]. For instance, Wang et al. [40] and Wang et al. [41] found enhanced drought caused the vegetation degradation in the southern and western regions, while Fang and Zhang [42] observed a negative impact of drought on vegetation in the southeastern areas. It was worth noting that there were larger discrepancies in previous studies on the effects of drought on vegetation on the TP. Therefore, the characteristics of drought and their effects on vegetation in the TP need to be further studied. Furthermore, whether the impacts of drought on vegetation have been consistent across different time periods in the past decades and how different vegetation types have responded to drought characteristics (frequency, duration, and intensity) were rarely addressed in previous studies, and this study will aim to answer these questions.

Drought was usually quantified through drought indices. Different types of droughts (meteorological, agricultural, hydrological, and socio-economic drought) had their own drought indices to characterize them. The commonly used meteorological drought indices included the Palmer Drought Severity Index (PDSI), the Standardized Precipitation Index (SPI), the Standardized Precipitation Evapotranspiration Index (SPEI), etc. [43]. Of the various drought indices, the SPEI not only accounted for the advantages of the PDSI, which took into account the effects of both P and potential evapotranspiration (PET) on drought, but also incorporated the advantages of the multi-timescale SPI [7,44,45]. Eventually, the SPEI has been broadly applied to study drought change and its impacts. In addition, the Normalized Difference Vegetation Index (NDVI) was a favored indicator for researchers to study vegetation change and its response to climate change [46,47].

This study showed the changes in meteorological drought and their impact on alpine vegetation in the growing season in 1982–2018 over the TP based on the SPEI and NDVI, with the aspects for the different drought impacts on vegetation over various time frames within the study period. Additionally, we investigated the impact of specific characteristics of drought on vegetation growth, including their effects on different vegetation types. The findings may further improve our comprehensive understanding of the impacts of drought on alpine ecosystems. The remainder of the article is structured as follows: Section 2 shows the study region, materials, and main research methods of the paper. The results of the paper are illustrated in Section 3. Section 4 and Section 5 present the discussion and main conclusions of the paper, respectively.

2. Materials and Methods

2.1. Study Region

The TP is located in the western region of China, which encompasses an area of approximately 2.57 × 10⁶ km², representing about 25% of the mainland of China. Its average elevation exceeds 4000 m [11,48]. The area of this study spanned from 75° to 105°E and from 25° to 40°N (Figure 1). As the highest plateau in the world, the TP has the world’s largest alpine ecosystem, comprising over 60% of the TP area covered by natural alpine grasslands, including alpine steppes and meadows [49,50]. The vegetation types over the TP include alpine shrubs, alpine sparse vegetation, croplands, and forests, in addition to the alpine grasslands, as presented in Figure 1b.

2.2. Materials

Observation stations on the TP are limited and primarily situated in the eastern region. This leads to a lack of in situ observations for studying climate change and its impact, as they cannot capture all signals across the entire TP, particularly in the broader northwestern region. Furthermore, calculating the SPEI requires multiple climatic elements, including near-surface wind speed (WIN), solar radiation (Rs), and relative humidity (RHU), in addition to Ta and P. Therefore, to conduct this study, a gridded dataset with complete variables and higher accuracy is necessary. We selected the China Meteorological Forcing Dataset (CMFD) to calculate SPEI due to its higher accuracy and complete variables. The dataset was available from the National Tibetan Plateau Data Center (https://www.tpdc.ac.cn, accessed on 1 March 2024), which comprises several climatic variables, including Ta, surface air pressure (Ps), specific humidity (q), WIN, surface downward shortwave radiation (rsds), surface downward longwave radiation (rlds), and surface precipitation rate. The spatial and temporal resolution of the dataset are 0.1° and 3 h, respectively, and spans from 1979 to 2018. The dataset was created by integrating data from the Princeton reanalysis, GLDAS, GEWEX-SRB radiance, TRMM P, and in situ observations from the China Meteorological Administration (CMA). The dataset’s accuracy was intermediate between satellite remote sensing and observation datasets, and was superior to that for internationally available reanalysis [51,52]. It was extensively evaluated and applied, and is considered suitable for the study of climate changes on the TP [53,54].

The NDVI data used in the paper were from the GIMMS NDVI dataset (GIMMS NDVI3g, V1.0) generated from AVHRR sensors. Maximum NDVI values were recorded within twice-monthly compositing periods (two values per month). Its spatial resolution was 0.083° and spans from 1982 to 2022 [55]. The GIMMS NDVI3g record extending from January 1982 to December 2018 in the growing season was used in this study. Previous researchers have verified that the dataset was highly suitable and accurate in investigating the changes in vegetation over the plateau areas [56].

The vegetation type data shown in Figure 1b were sourced from the 1:1 million China Vegetation Map, which was mainly based on the results of vegetation surveys conducted throughout the country, and comprehensively analyzed by combining the satellite imagery, aerial remote sensing, and relevant information on geology, pedology, and climatology. It can comprehensively reflect the geographical distribution of the 11 vegetation type groups, 833 groups and subgroups of 54 vegetation types, and about 2000 dominant species of communities, major crops, and economic plants in China. It was available at the National Cryosphere Desert Data Center (http://www.ncdc.ac.cn, accessed on 16 March 2024). The elevation data as shown in Figure 1a were sourced from the ERA-Interim with a spatial resolution of 0.75°, which could be accessed via the website https://www.ecmwf.int/en/forecasts/datasets/browse-reanalysis-datasets, accessed on 8 March 2024.

In order to harmonize spatial resolution of these datasets, we interpolated NDVI and each variable used to calculate SPEI to 0.05° using bilinear interpolation during our study. We also matched the vegetation type data to 0.05° by extracting the values from the nearest gridded points.

2.3. Methods

2.3.1. The Calculation of SPEI

The SPEI was computed on the basis of the deviation (D) of P and PET, representing the degree of dry–wet differences in a specific region [14]. The calculation process of SPEI was shown in Formulae (1)–(6) [57].

$$SPEI=W-\frac{B_0+B_{1}W+B_2{W}^2}{1+a_{1}W+a_2{W}^2+a_3{W}^3}$$

(1)

where

$$W=\sqrt{-2\ln \left(P\right)} \mathrm{f}\mathrm{o}\mathrm{r} P\le 0.5$$

(2)

Here, $P=1-F\left(x\right)$ represented the probability of exceeding a specified D value. If $P$ was larger than 0.5, it would be replaced with$1-P$ and the sign of the calculated SPEI would be reversed. In addition, B₀ = 2.515, B₁ = 0.802, B₂ = 0.010, a₁ = 1.432, a₂ = 0.189, and a₃ = 0.001. F(x) was the probability distribution function of the D series based on the Log-logistic distribution, which was calculated by

$$F\left(x\right)={\left[1+{\left(\frac{\alpha }{x-\gamma }\right)}^{\beta }\right]}^{-1}$$

(3)

where

$$\beta =\frac{2w_1-w_0}{6w_1-w_0-6w_2}$$

(4)

$$\alpha =\frac{\left(w_0-2w_1\right)\beta }{\mathsf{\Gamma }\left(1+1/\beta \right)\mathsf{\Gamma }\left(1-1/\beta \right) }$$

(5)

$$\gamma =w_0-\alpha \mathsf{\Gamma }\left(\frac{1+1}{\beta }\right)\mathsf{\Gamma }\left(\frac{1-1}{\beta }\right)$$

(6)

where $\mathsf{\Gamma }\left(\beta \right)$ was the gamma function of $\beta$. The SPEI in this paper was calculated using monthly P and PET through “SPEI” package in R language. The SPEI at 1-month, 3-month, and 6-month time scales were used in this study. The PET was computed by the Penman-Monteith formula using Ta, Ps, RHU, WIN, and radiation variables [58].

A drought event was identified when SPEI < −1.0 for one or more consecutive months in this study. Drought frequency (DF) was calculated as the number of droughts (n) divided by the total number of months (N) during the research period [59], as shown in Formula (7).

$$DF=\frac{n}{N}\times 100\%$$

(7)

Drought duration was described as the number of months that it persisted. Drought intensity was further calculated by Formula (8) [1]:

$$DI=\frac{\sum _{n=1}^T\left|S_{SPEI}-K\right|}{T}$$

(8)

where DI was the drought intensity, S_SPEI indicated the value of SPEI falling below the drought threshold (K), which was equal to −1 here, and T was the drought duration.

2.3.2. The Trend Analysis

The trend mutation of SPEI was investigated using the Mann–Kendall (M–K) test [60,61], which was extensively employed to detect the trend of hydrometeorological variables [62,63,64]. It was performed as follows:

For the sample time series X(x₁, x₂, x₃, …, x_n), an ordered sequence r_i was constructed to represent the cumulative number of samples S_k with x_i > x_j (1 ≤ j ≤ i) by

$$S_k={\sum }_{i=1}^k{r}_i\left(k=2, 3,\dots , n\right)$$

(9)

The value of r_i was calculated by

$$r_i=\left\{\begin{array}{c}1, x_i>x_j\\ 0, x_i\le x_j\end{array}\left(j=1, 2, \dots ,n\right)\right.$$

(10)

Assuming the time series was independent and random, the statistic ${UF}_k$ was calculated by

$${UF}_k=\frac{S_k-E\left(S_k\right)}{\sqrt{Var\left(S_k\right)}} \left(k=1, 2,\dots ,n\right)$$

(11)

where ${UF}_1=0$, E(S_k) and Var(S_k) represented the mean and variance of the sequence S_k, respectively. E(S_k) and Var(S_k) were calculated by

$$E\left(S_k\right)=\frac{n\left(n+1\right)}{4}$$

(12)

$$Var\left(S_k\right)=\frac{n\left(2n+5\right)\left(n-1\right)}{72}$$

(13)

$\left|{UF}_k\right|>U_{\alpha }$ represented significant trend changes in the sequence given the significance level of $\alpha$. ${UF}_k$ produced a curve $UF$. The same process was applied to the reverse sequence and assigned ${UB}_k=-{UF}_k\left(k=n, n-1, \dots , 1\right)$ and ${UB}_1=0$. The intersection of $UF$ and $UB$ curves represented an abrupt change.

Linear trends for SPEI and NDVI, as well as their relationship, were examined through the linear regression analysis and correlation coefficient, respectively. These calculations were performed with the “regCoef” and “escorc” functions in the NCAR Command Language (NCL).

2.3.3. The Extraction of NDVI

The NDVI was transformed from original semi-monthly to monthly data by selecting the maximum value of each month. Pixels with negative NDVI values that referred to snow and other contaminated data and pixels with poor quality were removed during the analysis of the paper.

In addition, the NDVI values of the pixels corresponding to each vegetation type according to the spatial distribution of each vegetation type including all alpine vegetation types, as shown on Figure 1b, were extracted. Thereafter, the effects of drought on the changes of each vegetation type were analyzed.

3. Results

3.1. The Spatio-Temporal Characteristics of SPEI

Figure 2a displays the interannual variation of the annual and growing season SPEI (1-month) averaged over the TP during 1982–2018. The annual SPEI averaged over the TP exhibits an insignificant rising trend (p > 0.05) at a change degree of 0.1/decade, indicating a wetting trend over the TP during the entire period. Additionally, the M–K test of the trend of the annual SPEI time series shows a change around 1995 (Figure 2b). Before 1995, the annual SPEI increases at an insignificant rate of 0.07/decade. However, after 1995, it decreases at a rate of −0.09/decade. The growing season SPEI averaged over the TP also shows a significant (p < 0.05) upward trend, with an average increase of 0.4/decade. Similar to the annual SPEI, the trend of the growing season SPEI also underwent a change around 1995 (Figure 2c). The growing season SPEI differs from the annual SPEI in that it shows a slight decreasing trend (–0.08/decade) before 1995 and an increasing trend (0.07/decade) thereafter. Gao et al. [65] reported that the aridity changes over the TP were closely related to P, which had a mutation around 1995 with a downward and upward trend before and after around 1995, respectively [66]. P on the TP is mainly concentrated in the growing season, so the SPEI in this period showed a decreasing trend before 1995 and an increasing trend after 1995. As for the annual SPEI, it may be affected by other factors such as Ta, which showed a more significant increasing trend after nearly 1995, which might induce an increase in the PET [67]. These trends suggest that the TP experienced wetting trends both annually and in the growing season during 1982–2018, but with different trends in various time periods. Previous findings based on observations also showed that the TP has become wetter in recent years, which was attributed to an increase in P [65,68].

Figure 3 presents spatial distribution of annual and growing season SPEI trends on the TP in 1982–2018. The annual SPEI presents an upward trend in most areas. The upward trend is primarily distributed in the northwestern regions, but the declining trend is mainly focused in the southern and eastern TP (Figure 3a). Therefore, the mean upward trend in the annual SPEI, as shown in Figure 2a, is largely contributed by the increasing SPEI trend in northwestern TP. The growing season SPEI increases over a wider range of areas and by a greater margin compared to the annual SPEI. The increasing trends pass the significance test at the 99.9% confidence level in most of the northern regions and the Yarlung Tsangpo River valley in southern regions of the TP (Figure 3b). These suggest that most regions of the TP have become wetter, except for parts of the southeastern and southern margins in 1982–2018. Gao et al. [65] and Zhang et al. [69] similarly stated a wetting trend in most of the stations on the TP, except for those in parts of the eastern region. The intensification of the local hydrological cycle and the enhanced transport of water vapor from the western Indian Ocean across India to the TP appeared to be the main reasons for the recent wetting of the TP [70].

Since 1995 is the critical year for the change in the SPEI (Figure 2), the spatial pattern of the SPEI trends for the periods of 1982–1995 and 1996–2018 are displayed separately in Figure 3c–f. In 1982–1995, the annual SPEI shows an upward trend in the majority regions on the TP, whereas declining trends are primarily concentrated in parts of the Qinghai Province and northwestern TP (Figure 3c). This ultimately leads to a mean upward trend in the annual SPEI averaged over the TP (Figure 2). Regarding the growing season SPEI, there is a wider range of the decreasing trend, particularly in Qinghai Province (Figure 3d). In 1996–2018, the spatial patterns of the SPEI trend both annually and for the growing season are significantly different from those before 1995 (Figure 3e,f). The annual SPEI reveals a significant decreasing trend in the majority of regions, particularly in the southeastern region (Figure 3e), resulting in a minor decline in the mean SPEI averaged over the TP as displayed in Figure 2a. In contrast to that before 1995, the growing season SPEI in 1996–2018 shows a declining trend in the southern regions and an upward trend in the northern regions (Figure 3f).

3.2. The Spatio-Temporal Characteristics of NDVI and Its Relationship with SPEI

The grassland on the TP shows a growing season from the end of May to October [71,72]. Therefore, to characterize vegetation changes more accurately, we will focus on the growing season when analyzing the NDVI. The spatial pattern of the annual NDVI in the growing season in 1982–2018 is shown in Figure 4a. It can be observed that the NDVI on the TP is characterized by high values in the southeastern regions and low values in the northwestern regions. The NDVI in the southeastern TP is higher than 0.65, but, in most regions of the northwestern TP, it is lower than 0.2. Figure 1b shows that the diversity of vegetation types over the TP is mostly concentrated in the southeastern regions, which is consistent with the distribution of larger NDVI values.

Annual NDVI averaged over the TP shows an insignificant (p > 0.05) upward trend at a rate of 0.0004/decade during 1982–2018 (Figure 4b). Spatially, the NDVI shows an upward trend in most parts of the northwestern TP. However, the NDVI displays a decreasing trend in the majority of the southeastern regions (Figure 5a), resulting in an insignificant increasing trend of the annual NDVI averaged across the entire TP, as presented in Figure 4b. These indicate that the vegetation of the TP sees greening during 1982–2018 when considering the TP as a whole. However, there is spatial heterogeneity, with greening mainly occurring in the western and northern regions, and degradation in the southeastern regions. Previous studies have also noted a greening trend in vegetation dynamics across the TP, with significant increases mainly observed in the northern region [34,73]. These findings further confirm the reliability of our results. As indicated in Section 3.1, the TP experienced a general tendency to be wet during the growing season, particularly in the northwestern area, and a drying trend in certain parts of the southeastern TP in 1982–2018 (Figure 3b). The spatial distribution of SPEI changes corresponds well with those of NDVI. Therefore, we can deduce that changes in wet and dry conditions have the potential to impact vegetation changes over the TP.

Figure 6 shows the correspondence linkage between the trends for the NDVI and SPEI in the growing season, further exploring the effect of changes in the SPEI on the NDVI. In the period of 1982–2018 (Figure 6a), 45% of the TP experiences a rising trend in the NDVI along with an increased SPEI trend, primarily in the northern TP. On the other hand, 12% of the TP located in the southeastern regions experiences a declining NDVI trend along with a declined SPEI. This indicates that the vegetation becoming better in most parts of the TP, especially in the northern and western parts, is largely due to the wetting trend, and the degradation of vegetation in most parts of the southeastern TP is induced by the drying trend in 1982–2018. Therefore, the areas where changes in the SPEI have an effect on the NDVI account for 57% of the TP. In approximately 43% of the TP, mainly in the southern regions, there is a poor relationship between NDVI and SPEI trends. This indicates that wetting and drying trends are not the main factor influencing vegetation growth in these regions.

Looking at different time frames, the mean annual NDVI averaged over the TP exhibits a significant upward trend with a change rate of 0.0087/decade before 1995. Spatially, the NDVI in most parts of the TP shows an upward trend during 1982–1995 (Figure 5b). The declining trend in the NDVI is largely observed in Qinghai Province situated in the northern TP. Section 3.1 findings indicate a declining SPEI trend in most of the northwestern regions and a rising trend in the southeastern regions in the growing season (Figure 3d). Figure 6b shows that regions representing 31% of the TP have an increase in the NDVI concurrently with an increase in the SPEI. These areas are mainly located in the southern regions of the TP. Conversely, approximately 15% of the TP has experienced a decrease in both the NDVI and SPEI. These regions are mainly distributed in Qinghai Province and Ali Plateau. Therefore, it can be concluded that the changes in vegetation cover on approximately 46% of the TP are due to alterations in wet and dry conditions during 1982–1995. However, the NDVI changes in regions covering 54% of the TP, mainly located in the western and northern TP, are not primarily influenced by changes in the SPEI as they do not show consistency with each other. The factors that induce vegetation greening in this period need further discussion.

After 1995, the mean annual NDVI averaged over the study area decreases at a rate of −0.0049/decade. Pang et al. [74] also pointed out that the NDVI had a significant upward trend until the late 1990s, and, since then, it has shown a slightly downward trend. The spatial pattern of the NDVI trend during 1996–2018 shows an upward trend in the northern regions and a downward trend in the southern regions. In some areas, the decreasing trend passes the significant test (Figure 5c), resulting in the decreased NDVI averaged over the TP (Figure 4b). Figure 3f shows an increase in the SPEI in the northern TP and a decrease in the southern regions. Therefore, drying and wetting trends significantly affect the vegetation changes during 1996–2018. According to Figure 6c, approximately 23% of the TP’s total area shows an increase in the NDVI along with an upward SPEI, mainly located in the northern TP. Moreover, about 38% of the TP’s area, mainly in the southern TP, shows a decreased NDVI along with a decreased SPEI. Consequently, the vegetation in the area, accounting for a total of 61% of the study area, is influenced by the changes in dry and wet conditions in 1996–2018.

3.3. The Impacts of Drought Characteristics on Different Vegetation Types

In addition to the drying and wetting trend, drought characteristics such as the frequency, duration, and intensity can also impact vegetation. Figure 7 shows the frequency of drought that occurred in 1982–2018, calculated based on the SPEI at different timescales. Based on the 1-month-timescale SPEI, droughts occur more frequently in the eastern parts of the TP than in the western parts (Figure 7a). The spatial pattern of the drought frequency in the growing season is similar to that of the annual drought frequency. The Qilian Mountain region and the confluence of the Yangtze, Salween, and Mekong Rivers sources are the major centers of high drought frequency, where the frequency of drought can exceed 17% (Figure 7b). The spatial patterns for the drought frequency derived from the 3-month- and 6-month-timescale SPEI are similar to that of the 1-month-timescale SPEI, with a higher frequency in the eastern TP and a lower frequency in the western TP (Figure 7c,e). Moreover, the drought frequency in the growing season (Figure 7d,f) is higher than that annually. Additionally, as the SPEI timescale increases, there is a corresponding decrease in drought frequency. The frequency of droughts in the western TP, as determined by the 6-month-timescale SPEI, is almost less than 5% (Figure 7e).

The mean duration of the drought that occurred during 1982–2018 is shown in Figure 8. It shows that the spatial pattern of the drought duration based on the 1-month timescale is generally inversely distributed with the spatial pattern of the drought frequency. This means that areas that have high drought frequency values correspond to those with low drought duration values. The mean drought duration values are larger in the northern side of Kunlun Mountain, where they exceed 2 months (Figure 8a). The duration of drought during the growing season is shorter than that annually, especially at the eastern parts of the TP (Figure 8b). The spatial patterns of the drought duration based on 3-month- and 6-month-timescale SPEI are similar to that based on the 1-month-timescale SPEI. The longer duration occurs in the western regions, while the shorter duration occurs in the eastern regions of the TP (Figure 8c,e). Additionally, the drought durations during the growing season are shorter than those annually (Figure 8d,f). Unlike the drought frequency, the mean duration of drought increases as the SPEI timescale increases. Based on the 6-month SPEI (Figure 8e), most regions of the TP experience drought for a duration of over 3 months.

As for the drought intensity, the spatial distribution pattern across the TP is similar to that for the drought duration. According to the results of the 1-month-timescale SPEI, droughts that occur in the northwestern regions are more intense those in the southeastern TP, and the most severe droughts occur mainly in the northwestern TP (Figure 9a). Moreover, the annual drought intensity is stronger than that during the growing season (Figure 9b). The distribution characteristics of the drought intensities based on the 3-month- and 6-month-timescale SPEI are similar to those based on the 1-month-timescale SPEI. As the SPEI timescale increases, the severity of drought also increases. On the 6-month timescale, the majority of regions in the TP experience a drought intensity above 6 (Figure 9e).

The relationships between the drought characteristics and the NDVI of various vegetation types are displayed in Figure 10. Generally, vegetation condition deteriorates as droughts become more frequent, longer in duration, and more intense. Figure 10(a1–c1) show that the NDVI has a positive correlation with drought frequency, with correlation coefficient of 0.57, but a negative correlation with drought duration and intensity, with correlation coefficients of −0.40 and −0.52, respectively. Therefore, drought duration and intensity pose a greater influence on the spatial pattern of the NDVI across the TP, with drought intensity having a more pronounced effect due to its higher spatial correlation coefficient of −0.52 (Figure 10(c1)). When examining various vegetation types, it is found that, with the exception of forests (Figure 10(a5–c5)) and croplands (Figure 10(a7–c7)), all other types of vegetation are positively correlated with drought frequency, with positive correlation coefficients, and negatively correlated with drought duration and intensity, with negative correlation coefficients. Therefore, the spatial distributions of the NDVI for all vegetation types, except for croplands and forests, are primarily influenced by the drought duration and intensity, rather than their frequency. The effect of drought intensity on the NDVI is most significant for sparse vegetation, with a correlation coefficient of −0.53 (Figure 10(c2)). Meadow (Figure 10(c3)) and grassland (Figure 10(c6)) follow, with correlation coefficients of −0.43 and −0.31, respectively. Drought duration also has a substantial impact on sparse vegetation (Figure 10(b2)) and meadow (Figure 10(b3)); the correlation coefficients between them are −0.44 and −0.43, respectively, followed by shrub (Figure 10(b4)). The spatial distribution of the NDVI for forests is less affected by drought characteristics, which may be attributed to their more developed root systems. Forests can significantly mitigate the drought influence by drawing on available groundwater resources. In addition, drought intensity has some impact on cropland (Figure 10(c7)). However, it may be because of the significant impact of human activities on cropland, such as irrigation; its spatial correlation coefficients with drought characteristics are not very high.

4. Discussion

In this study, we investigated the characteristics of drought change and its relationship with vegetation change on the TP using the SPEI and NDVI. The results show that changes in the NDVI are strongly correlated with changes in the SPEI. However, in some time periods and regions, vegetation changes cannot be solely explained by drought changes. For example, the SPEI shows a decreasing trend, while the NDVI shows an upward trend over large parts of the TP from 1982 to 1995. Therefore, the vegetation change during this period could be influenced primarily by other factors, such as snow cover [75] and permafrost [76]. The regions where the NDVI and SPEI are not synergistic are mainly located in the northward part of the TP. It is worth noting that the TP contains a significant amount of snow [77] and glacier [78] due to its high elevation topography and climatic features. Additionally, the TP has permafrost, accounting for 75% of the permafrost in the mountainous regions of the Northern Hemisphere [79]. These cryosphere features are mainly distributed in the northern TP [80], which is consistent with the distribution of inconsistent changes in the NDVI and SPEI. Furthermore, the northern region of the TP showed a higher degree of warming than the southern part from 1982 to 1998 [68,81]. The melting of snow and glaciers resulting from rapid warming can replenish water during droughts through regulating the soil moisture, which reduces the hydraulic stress on vegetation growth. The warmth-induced increases in the thickness of the active layer and the thawing days of permafrost can also improve soil water, which may alleviate the phenomenon of water stress on the growth of vegetation [82]. These could temporarily promote vegetation growth. Therefore, the impact of cryosphere changes due to climate warming may have contributed to the increase in vegetation greening over most regions of the TP in 1982–1995, as shown in our results. Many other works of research also demonstrated the significant roles that snow cover, glaciers, and permafrost played in vegetation changes under climate warming [76,83,84]. In addition to the factors mentioned above, the land surface temperature [7,85] and other non-climatic factors [86], such as a CO₂ fertilization increase [87], terrain [88], soil properties [89] with varying water retention capacities due to differing soil textures [90], and grazing, will also affect the growth and coverage of vegetation.

The analyses in this paper were conducted at a monthly timescale for both the SPEI and NDVI; they can be conducted on smaller timescales, such as daily or hourly, to obtain a more detailed description of drought variation and its effects on vegetation in future research. Furthermore, the meteorological variables utilized in the calculation of the SPEI are sourced from the CMFD dataset, which is more accurate than other datasets. However, its resolution is limited to 0.1° and the time frame only spans from 1979 to 2018. This resolution may not be sufficient for the TP, given its complex topography, as it may not capture more detailed climate change information. Therefore, a higher resolution and longer time series of gridded observational datasets are critically needed to study drought characteristics and their impacts on the vegetation over the TP.

5. Conclusions

The study explores the characteristics of drought changes and their impacts on vegetation during the growing period using the SPEI and NDVI over the TP in 1982–2018. The main findings are as follows:

The majority of regions over the TP show a wetting trend over the study period, particularly in the growing season, with a mean change rate of 0.4/decade in the SPEI. Additionally, the SPEI trend in the growing season has a change around 1995. Before 1995, there is a drying trend in the northern regions and a wetting trend in the southern regions of the TP, with an SPEI change rate of −0.08/decade. However, after 1995, the wetting trend occurs in the northern regions and the drying trend appears in the southern regions, with an SPEI change rate of 0.07/decade.

The NDVI results show that the overall vegetation condition across the TP showed a positive trend in 1982–2018. Vegetation changes on the TP are dominated by wetting and drying trends throughout the study period and after 1995, in total, affecting 57% and 61% of the TP, respectively. However, during 1982–1995, the NDVI presents a significant upward trend over larger parts of the TP with an increase rate of 0.0087/decade, while the SPEI shows a decreasing trend. Therefore, the drying trend is not the influencing factor of vegetation growth during this period.

Droughts in the northwestern regions of the TP are longer and more severe than in the southeastern regions. Conversely, droughts occur more frequently in the southeastern regions than in the northwestern. The spatial pattern of the NDVI is closely correlated to drought duration and intensity, particularly to drought intensity, with a spatial correlation coefficient of −0.52. The spatial distribution of the NDVI for sparse vegetation is most significantly affected by drought intensity, with a correlation coefficient of −0.53.