Altitude-Shifted Climate Variables Dominate the Drought Effects on Alpine Grasslands over the Qinghai–Tibetan Plateau

Abstract

Drought has broad and deep influences on ecosystem dynamics and functions, particularly considering the lagged and cumulative effects of drought. Yet the individual role of climate variables in mediating such drought effects on vegetation remains largely unknown. Based on the Normalized Difference Vegetation Index (NDVI) and the standard precipitation evapotranspiration index (SPEI), here, we investigated the patterns and mechanisms of drought effects on alpine grasslands in the Qinghai–Tibetan Plateau (QTP) from 1982 to 2015. Drought imposed widespread lagged and cumulative impacts on alpine grasslands with notable spatial heterogeneity, showing that the southwestern and northeastern parts of the plateau were more sensitive and responded quickly to drought. Further, drought effects showed an evident elevation dependence across different grassland types, which could be explained by altitudinal shifts in climatic factors, including temperature and precipitation. Precipitation was the dominant factor in drought effects on alpine meadows, while temperature dominated the drought impacts on the alpine steppes. Such a divergent dominant factor implied that there would be different vegetation responses to future climate change among diverse types of alpine grasslands. To maintain the sustainability of alpine grassland, more effort should be applied to alpine steppes regarding pasture management, particularly in response to extreme drought due to warmer climates in the future.

1. Introduction

In the context of ongoing climate change, climate extremes have caused severe impacts on the structure and function of terrestrial ecosystems [1,2,3,4]. As one of the most devastating extreme weather events, drought imposes a far-reaching influence on ecosystem productivity [5,6,7]. Drought can significantly reduce vegetation productivity by constraining vegetation growth [8,9], causing massive plant mortality [10,11,12]. Additionally, the drought-related risks of reducing vegetation productivity are expected to increase under future warming [13,14]. Vegetation provides a lot of ecosystem services, such as food production and carbon uptake, and plays a crucial role in maintaining ecosystem sustainability. An in-depth understanding of the vegetation response to drought is thus critical to evaluating the current sustainability of ecosystems in response to climate extremes and predicting it under future climate changes.

Generally, drought chronically affects vegetation growth [15,16]. For example, current vegetation dynamics might be driven by the impacts of previous water stress due to the lagged effects of climate variables [17,18]. Moreover, drought can also impose a cumulative effect on vegetation performance since it spans a period of time, leading to persistent water constraints in plants [19,20,21]. Combining such lagged and cumulative effects enables a comprehensive assessment of drought impacts on vegetation. Previous studies have demonstrated that the magnitude of lagged and cumulative effects vary among the water conditions typically represented by the annual standard precipitation-evapotranspiration index (SPEI) [22,23,24,25,26]. Yet the individual role of temperature in mediating drought effects has received less attention. Temperature is a crucial factor in vegetation growth and regulates the vegetation response to external disturbance [27]. Growing evidence has also documented how the response of ecosystems to drought could change a lot under warmer environments [28,29]. Therefore, quantifying the potential contributions of temperature to the variability of drought effects is also imperative to comprehend the underlying mechanisms in the vegetation response to drought.

As a key ecological security barrier, the Qinghai–Tibetan Plateau (QTP) is crucial for regional climate and global energy–water cycles [30,31,32]. At the same time, the QTP is a vulnerable region in regard to ongoing climate change due to a typical highland climate [33,34]. Recently, large-scale and rapid warming (roughly twice the global average) has been observed on the QTP [35,36,37,38]. The evident warming has extended the vegetation growth season length and enhanced vegetation productivity over the QTP [39,40]. Nevertheless, the increasing temperature has also accelerated evaporation, exacerbated water deficiency, and thereby exerted damaging impacts on vegetation growth [41,42]. For instance, several studies have shown that warming triggers the substantial browning of vegetation by reducing water availability in the middle and southwestern QTP [43,44]. Hence, considering the rapidly changing climate, drought is projected to trigger more negative influences on vegetation productivity on the QTP against a warming background [45,46].

Covering roughly two-thirds of the total plateau area, alpine grasslands are the major vegetation types on the QTP and maintain fundamental ecosystem services (such as food production) to local human society [47,48]. Alpine grasslands on the QTP can be classified into diverse types according to their dominant species. Among them, the most widely distributed grassland types are alpine meadows and steppe [49,50]. Alpine meadows are characterized by relatively abundant water availability, while the alpine steppe has suffered from limited precipitation [51,52]. Meanwhile, climate variables, mainly including temperature and precipitation, display obvious elevation-dependent characteristics on the QTP [53,54,55]. Further, the alpine grasslands’ sensitivity to climate change also varies with elevation gradients [56,57,58]. Here, the two widespread grassland types and huge elevational ranges on the QTP provide an opportunity to explore the potential drivers of drought effects on vegetation growth.

In this study, leveraged by the Normalized Difference Vegetation Index (NDVI) and the corresponding standard precipitation evapotranspiration index (SPEI), we calculated drought effects, including lagged and cumulative effects, on QTP grasslands and further investigated their variations along the gradients of elevation and climate variables. Our overall objectives are as follows: (1) to detect the spatial distributions of drought effects on alpine grasslands; (2) to reveal how these effects change along elevation gradients and, further, to identify the predominant climatic contributor.

2. Materials and Methods

2.1. Study Area

The QTP is located in southwest China and has a unique environment with low temperatures, intensified solar radiation, limited precipitation, and complex terrain [59,60]. The elevation of alpine grasslands ranges from less than 2500 m to more than 5500 m (Figure 1a,b). The mean temperature of the growing season from 1982 to 2015 in the grasslands increased from less than 0 °C to greater than 12 °C (Figure 1c). The total precipitation of the growing season ranges from 100 mm to 500 mm (Figure 1d). The growing season is limited to May to September of each year.

2.2. Data

2.2.1. NDVI Dataset

To capture the vegetation dynamics, we used the NDVI as the indicator of vegetation growth. The GIMMS3g NDVI for 1982–2015 was derived at https://poles.tpdc.ac.cn/en/data/9775f2b4-7370-4e5e-a537-3482c9a83d88/, accessed on 15 May 2024 [61]. This biweekly NDVI dataset is one of the longest time series of satellite-based vegetation indicators with an 8 km × 8 km grid resolution. This NDVI dataset minimized the influence of sensor degradation, cloud cover, and volcanic aerosols and showed good performance in temporal consistency. To eliminate the impacts of sparsely vegetated areas, grid cells with a mean NDVI for growing season less than 0.1 were excluded from the analysis. To match the resolution of the climate dataset, the NDVI dataset was resampled to 0.1° grid resolution by adopting the nearest-neighbor algorithm and aggregated into a monthly timescale via maximum value composite.

2.2.2. Climate Data

The monthly mean temperature and total precipitation were obtained from the China meteorological forcing dataset (CMFD) (http://data.tpdc.ac.cn/en/, accessed on 15 May 2024), which was developed by fusing the climate variables of satellite-based observations, reanalysis datasets, and meteorological station data [62]. With a grid resolution of 0.1°, the CMFD is generally better at evaluating accuracy than other climatic products in China.

2.2.3. Vegetation Cover and Elevation Dataset

The grassland types on the QTP were derived from the 1:1,000,000 vegetation atlas of China (https://poles.tpdc.ac.cn/zh-hans/data/eac4f2cf-d527-4140-a35d-79992957f043/, accessed on 15 May 2024), which was completed by more than 200 scientists after more than 30 years of work and has been widely used in previous ecological studies. The vegetation map shows the distribution of 868 basic vegetation taxonomic units in China. It shows, in detail, the distribution and geographical pattern of vegetation, including horizontal distribution, vertical distribution, and its relationship with climatic factors and ground environmental factors. The elevation (spatial resolution of ~1 km) of the Global 30 Arc-Second Elevation Dataset (GTOPO30, https://www.usgs.gov/centers, accessed on 15 May 2024) was downloaded from the Google Earth Engine (GEE) platform. GTOPO30, completed in late 1996, is a global digital elevation model (DEM) derived from several raster and vector sources of topographic information. To match the spatial resolution of the climate data, both the vegetation cover and elevation dataset were resampled to a 0.1° grid resolution by adopting bilinear interpolation.

2.3. Method

2.3.1. Drought Index

The standard precipitation evapotranspiration index (SPEI), derived from the deviations from average climatic water balance (Equation (1)), is widely used in the assessment of the intensity and duration of drought [63]. The SPEI synthesizes the effects of temperature variability into drought evaluation and has the capacity to identify droughts at different scales [19,64]. The calculation method for climatic water balance is as follows:

$$B_i=P_i-{PET}_i$$

(1)

where i is the month, and $B_i$, $P_i$, and ${PET}_i$ are the monthly water balance (mm), the monthly total precipitation (mm), and the monthly potential evapotranspiration (mm), respectively.

To obtain the SPEI values, we first calculated the gridded potential evapotranspiration following the method of Thornthwaite [65] as follows:

$${PET}_i=16\times {\left({\displaystyle \frac{10T_i}{H}}\right)}^A\times \left({\displaystyle \frac{N}{12}}\right)\times \left({\displaystyle \frac{D}{30}}\right)$$

(2)

$$H={\displaystyle {\sum }_{i=1}^{12}}{\left({\displaystyle \frac{T_i}{5}}\right)}^{1.154}$$

(3)

$$A=0.49239+1792\times {10}^{-5}H-771\times {10}^{-7}{H}^2+675\times {10}^{-9}{H}^3$$

(4)

where $T_i$, A, N, D, and H are the mean temperature, constant, maximum sunny hours, number of days per month, and heat index, respectively. It should be noted that the N is estimated by the latitude.

Then, we normalized the monthly aggregated water balance from Equation (1) at multiple timescales (from 1 to 24 months) to obtain the SPEI values (http://spei.csic.es/home.html, accessed on 15 May 2024). The values of the SPEI indicate climatic water availability, with larger positive values suggesting more water surplus and vice versa. Moreover, the SPEI can represent the previous cumulated water availability with multiple timescales. For instance, the values of the 1-month and 3-month SPEIs indicate the water conditions of the current month and the previous three months, respectively.

2.3.2. Lagged and Cumulative Effects of Drought on Vegetation

The lagged and cumulative effects were calculated based on the maximum Pearson’s correlation coefficient between NDVI and SPEI [19,21,22]. Specifically, for the lagged effect, the correlation coefficients between the NDVI and 1-month SPEI among the preceding 24 months were first calculated. For the cumulative effect, the correlation coefficients were derived from the NDVI and the corresponding different timescales (i.e., from 1 to 24 months) of the SPEI. Then, the lagged month-of-drought effect was determined, in which the preceding month where the maximum correlation coefficient between the NDVI and the 1-month SPEI was observed. The accumulated month-of-drought effect was determined, in which the timescale of the SPEI where the maximum correlation coefficient between the NDVI and the SPEI is shown. Meanwhile, the maximum correlation coefficients indicated the strength of the lagged and cumulative effects. Generally, the shorter drought timescales (i.e., lagged and accumulated times) imply that vegetation reacts more quickly to drought, indicating higher sensitivity [66,67].

$$R_i=corr\left(NDVI,{SPEI}_i\right)\hspace{1em}1\le i\le 24$$

(5)

$$R_{\mathrm{m}\mathrm{a}\mathrm{x}\_lag}=\max \left(R_i\right)\hspace{1em}1\le i\le 24$$

(6)

where $R_i$ represents the correlation coefficient at the lagged time of the $i$ month, and ${SPEI}_i$ and $R_{\mathrm{m}\mathrm{a}\mathrm{x}\_lag}$ denote the 1-month timescale SPEI at previous i month and the maximum value of $R_i$, respectively.

$$R_j=corr\left(NDVI,{SPEI}_j\right)\hspace{1em}1\le j\le 24$$

(7)

$$R_{\mathrm{m}\mathrm{a}\mathrm{x}\_cml}=\max \left(R_j\right)\hspace{1em}1\le j\le 24$$

(8)

where $R_j$ represents the correlation coefficient between the NDVI and SPEI at a j-month timescale. ${SPEI}_j$ and $R_{\mathrm{m}\mathrm{a}\mathrm{x}\_cml}$ denote the j-month timescale SPEI and the maximum value of $R_j$, repectively.

2.3.3. Elevation and Climate Gradients

To investigate the elevation and climate gradients of drought effects in grassland vegetation, we calculated the mean lagged and cumulative effects in bins of elevation at an interval of 100 m. Then, the average temperature and total precipitation were summarized in each elevation bin. When calculating drought effects along elevation, only the grid cells with significant correlation coefficients between the NDVI and the SPEI were included, and elevation bins with a sample size of less than 100 were removed.

2.4. Statistical Analysis

Linear or quadratic regression was adopted to estimate the relationships between drought effects and elevation, as well as climatic variables. It should be noted that we chose the above two regression methods due to their explicit ecological implications. For example, the linear regression between temperature and drought effects implies that drought effects increase or decrease along with temperature, while quadratic regression suggests that there is an optimum temperature. Then, the Akaike information criterion (AIC) was used to determine the best regression between linear and quadratic fitting, with the lower values suggesting better performance. Finally, we employed the random forest algorithm (RF) [68] to evaluate the relative contributions of elevation and climatic variables (temperature and precipitation) to the effects of variation in drought effects on grassland vegetation. We used 70% of the data for model training and 30% for model validation. Finally, we built an RF model with 500 regression trees. In the RF, the mean square error (%IncMSE) represents the importance of each predictor (i.e., higher values indicate higher variable importance) (Ho 1998).

All calculations and statistical tests were conducted in R software version 4.3.0 [69]. The “SPEI” package was used to calculate the SPEI value, the “Hmisc” package to calculate Pearson’s correlation, the “stats” package to conduct linear and quadratic regression, and the “randomForest” package to create a random forest model.

3. Results

3.1. Spatial Patterns of Drought Effects

For the lagged effects, about 99.6% of grasslands presented positive correlations, and 54.52% of those were statistically significant (p < 0.05), primarily distributed in the northeastern and southwestern parts of the plateau (Figure 2a). Among the pixels showing lagged effects, about 40.1% of them exhibited shorter lagged times (range of 1 to 4 months), mainly located in the western and northeastern parts of the plateau (Figure 2b). Conversely, approximately 12.2% of the areas showed longer lagged times (range of 21 to 24 months) and were concentrated on the central plateau.

For the cumulative effects, most grassland areas (80.28%) in the QTP were dominated by positive correlation, while 19.71% of areas showed negative correlations (Figure 3a). In addition, 32.01% of grasslands, mostly located in the southwestern and northeastern parts of the plateau, exhibited significantly positive correlations (p < 0.05). Negative correlations, despite almost none meeting the significance criteria, were primarily observed in the northwestern and southeastern parts of the plateau. Considering the accumulated timescale, the short-term (1–4 months) accumulated months principally occupied the southwest and northeast plateau (covering 12.20% of the total plateau areas), whereas medium (13–16 months) and long-term (21–24 months) accumulated months mostly occurred in the central plateau (amounting to 23.30% and 30.27% of total plateau areas, respectively) (Figure 3b).

3.2. Variation in Drought Effects along the Gradients of Elevation, Temperature, and Precipitation

For alpine meadows, a parabolic-shaped relationship was observed between lagged months and elevation (p = 0.001) and a U-shaped relationship between R_{max_lag} and elevation (p = 0.001) (Figure 4a; Supplementary Figure S1a). The lagged months first increased at altitudes below 4300 m and then decreased at altitudes greater than 4300 m. However, there are no relationships between accumulated months/R_{max_cml} and elevation (Figure 4c, Supplementary Figure S1c). For the alpine steppe, the lagged (p < 0.001) and accumulated (p < 0.001) months were significantly positively correlated with elevation (Figure 4b,d). Additionally, R_{max_lag} (R² = 0.72, p < 0.001) and R_{max_cml} (R² = 0.72, p < 0.001) decreased significantly along the increasing elevation. These results suggest that the alpine steppes are more susceptible and respond more quickly to drought at lower altitudes (Supplementary Figure S1b,d).

For alpine meadows, a parabolic-shaped relationship was observed between lagged months and temperature (p = 0.001) and a U-shaped relationship for R_{max_lag} and temperature (p = 0.001) (Figure 5a; Supplementary Figure S2a). The lagged months first increased at temperatures below 6.1 °C and then decreased at altitudes greater than 6.1 °C. However, the accumulated months and R_{max_cml} did not show significant linear or quadratic changes along the temperature gradients (Figure 5c; Supplementary Figure S2c). For the alpine steppe, the lagged (p < 0.001) and accumulated (p < 0.001) months were significantly negatively correlated with temperature (Figure 5b,d). Meanwhile, R_{max_lag} (p < 0.001) and R_{max_cml} (p < 0.001) significantly increased along with increasing temperatures. Both of these findings indicate that alpine steppes are more sensitive and respond more rapidly to drought in hotter areas (Supplementary Figure S2b,d).

For the alpine meadow, the lagged months were significantly positively correlated with precipitation during the growing season (p < 0.001), while R_{max_lag} was significantly negatively correlated (p < 0.001) (Figure 6a; Supplementary Figure S3a). Meanwhile, a parabolic-shaped relationship was exhibited between the accumulated months and precipitation (p < 0.01), while there was a U-shaped relationship for R_{max_cml} and precipitation (p < 0.05) (Figure 6c, Supplementary Figure S3c). The accumulated months first increased at temperatures below 385 mm and then decreased at altitudes greater than 385 mm. For the alpine steppe, the lagged months, accumulated months, R_{max_lag}, and R_{max_cml} showed no significant relationships with precipitation (Figure 6b,d; Supplementary Figure S3b,d).

3.3. Divergent Contributions of Elevation, Temperature, and Precipitation to Drought Effects

Overall, the random forest model showed good performance in simulating the observed variables regarding drought effects on the alpine meadow and steppe (Supplementary Figures S4–S6). As for alpine meadows, elevation contributed the most variations in lagged effects and accumulated months (Figure 7a,c). Among the climatic predictors, precipitation contributed more variations in accumulated months than temperature. For the alpine steppe, elevation contributed the most variations in lagged and accumulated months (Figure 7b,d). Among the climatic predictors, temperature was a more important variable than precipitation. In addition, the results of R_{max_lag} and R_{max_cml} were similar to the findings for the lagged and accumulated months (Supplementary Figures S5 and S6).

4. Discussion

4.1. The Spatial Patterns of Drought Effects on Alpine Grassland

Overall, alpine grasslands were dominated by positive correlation coefficients in regard to the lagged and cumulative effects, indicating vegetation growth on grasslands is substantially influenced by previous climatic water conditions (Figure 2a and Figure 3a). The underlying mechanisms may be related to the unique climatic environment of alpine grasslands and the functional traits of herbaceous plants. First, alpine grasslands mainly belong to arid and semi-arid regions [70]. As they are characterized by infrequent and highly variable precipitation, water availability is a critical factor in affecting vegetation growth in alpine grasslands [71,72]. Second, the grasses merely capture the water from the surface of soil layers due to their relatively short and shallow roots [73,74]. Such root characteristics prevent the herbs from absorbing the deep soil water to escape drought stress, leading to a strong relationship between vegetation growth and water availability in alpine grasslands.

Additionally, we found that the drought effects were spatially heterogeneous. On the one hand, the significant positive correlation coefficients were primarily distributed in the southwestern and northeastern parts of the plateau (Figure 2a and Figure 3a). On the other hand, the short-term (1–4 months) lagged/accumulated months principally occupied the southwestern and northeastern parts of the plateau, whereas medium (13–16 months) and long-term (21–24 months) lagged/accumulated months mostly occurred in the central plateau (Figure 2b and Figure 3b). These results indicated that the southwestern and northeastern parts of the plateau are more susceptible and respond more quickly to climatic water deficits. This phenomenon may be attributable to the distribution patterns of different grassland types. Spatially, alpine steppes are largely aggregated in the western and northern parts of the QTP, while alpine meadows are mainly distributed in the central and eastern parts of the plateau [70]. Alpine steppes generally exhibit higher vulnerability to drought stress than alpine meadows [29].

4.2. Divergent Variation in Drought Effects along Elevation and Climate Gradients among Different Grassland Types

Generally, the drought effects on alpine grasslands varied with elevation, which can be attributed to altitudinal shifts in climate variables (e.g., temperature and precipitation). For alpine meadows, the magnitude of lagged months first increased at lower altitudes (below 4300 m) and then decreased at higher altitudes (greater than 4300 m) (Figure 4a). This phenomenon was consistent with the finding that the drought effect generally increased with increasing elevation but showed a decline when the altitude ranged from 3000 to 5000 m in the global mountain areas [21]. Meanwhile, the variation in lagged months exhibited a parabolic-shaped relationship with temperature gradients and a positive relationship with precipitation (Figure 5a and Figure 6a). Furthermore, the random forest results imply that water availability is the dominant factor of lagged months. These results suggest that alpine meadows are sensitive to drought in more arid conditions (Figure 6a). Sufficient precipitation could serve as a buffer by increasing the soil moisture content for vegetation growth when drought occurs [20,75]. Meanwhile, the coupling of vegetation dynamics and water availability could be strengthened when ecosystems experience declined precipitation [66,76]. Therefore, the magnitude of drought effects on alpine meadows varies along with the water conditions, which was also corroborated by previous studies [23,26]. For instance, previous studies have revealed that the drought effects on global grasslands were stronger in arid regions than in humid areas. Meanwhile, there is also numerical evidence that global vegetation tends to be more tolerant to drought in humid ecosystems.

For the alpine steppe, drought effects showed linear altitude dependencies, as demonstrated by lagged and accumulated months rapidly increasing along elevation gradients (Figure 4b,d). These results suggested that alpine steppes are more susceptible to water deficits and respond more quickly to short-term drought at lower altitudes, which is in line with previous evidence showing that drought effects decrease with increasing elevation [29,77]. At the same time, positive relationships were found in the variation in lagged and accumulated months along with increasing temperature, indicating a more pronounced drought effect on alpine steppes under warmer conditions (Figure 5b,d). This result was in contrast to previous studies that typically revealed that water availability, not temperature, is the dominant factor affecting the vegetation response to drought. Our study object, that is, the alpine steppe, is the primary reason for this contrasting result. Compared with the other ecosystems investigated in previous studies, the alpine steppe is more likely to suffer from the limitations of energy and temperature. Thus, temperature plays a critical role in mediating vegetation productivity and, therefore, the vegetation response to drought. A recent study found that the magnitude of vegetation growth responses to drought is mainly influenced by temperature in the QTP, particularly at 3000–4500 m intervals [29,78]. Moreover, it has also been documented that drought accompanied by warmer temperatures causes more damage to plants compared with simple water deficits [79,80,81]. Additionally, there were non-significant relationships between drought effects and precipitation (Figure 6b,d). Largely located in hyper-arid and arid areas, alpine steppes suffer from chronic water stress [82]. We speculate that the physiological and functional characteristics of plants in alpine steppes may evolve to adapt to the arid environment, leading to high resilience and resistance to precipitation variations [83,84]. Finally, the results of the random forest model also showed that temperature is a more important variable than precipitation for drought effects (Figure 7b,d). Therefore, against the background of rapid warming, the alpine steppe is expected to be more vulnerable to water deficiency since increasing temperatures will shorten its response time to drought.

4.3. Limitations and Implications for Future Studies

Here, we detected the spatial patterns of drought effects on alpine grassland and further analyzed how elevation and climatic variables affect such drought effects among various grassland types. Specifically, we found that precipitation was the dominant factor in drought effects on alpine meadows, whereas temperature is the main driving factor for the responses of alpine steppe vegetation to drought. Such a divergent dominant factor implies that there will be different vegetation responses to future climate change among the two types of alpine grassland. Considering that the QTP is projected to become warmer and wetter, drought effects on vegetation could be exacerbated in the alpine steppe due to elevated temperatures and be alleviated in alpine meadows because of increased precipitation. Thus, our study highlighted that we should pay more attention and apply more effort to the alpine steppe in response to future droughts. Nevertheless, our study is still encumbered by some limitations and uncertainties, which hamper our comprehensive understanding of the vegetation response to drought. First, the spatial heterogeneity of the vegetation growth response to drought in different elevation intervals needs further exploration, as there are varying hydrothermal conditions at various elevations. Multiple drought indexes should be incorporated to enhance the robustness of the results, for example, the normalized Reconnaissance Drought Index (RDI). It was shown that there was no significant influence on RDI results detected by four popular PET methods (Hargreaves, Thornthwaite, Blaney–Criddle, and FAO Penman–Monteith (only T)), regardless of the reference period analysis. Second, we argue that the apparent spatial heterogeneity in drought effects on the alpine grasslands was associated with different hydrothermal conditions and diverse functional strategies for different grassland types. Other environmental factors, such as soil properties and biodiversity [85,86,87], may also alter the vegetation response to drought. For instance, the high species diversity and complex community structure could enhance resistance to drought stresses, as illustrated by forests with higher species diversity that are less affected by changing water conditions [88,89,90]. Additionally, we also did not take the impact of ecological disturbances on vegetation growth into consideration. Ecological disturbances, including land use changes and human activities [35,59], may pronouncedly affect the growth of vegetation and thereby reshape the influence of drought on the alpine grasslands. Thus, future studies may account for more environmental indicators to identify potential control factors in shaping drought effects. Third, due to the limited large-scale control experiments conducted on the QTP, we employed statistical methods (including regression analysis and random forest arithmetic) to explore the relationship between drought effects and their possible driving forces. However, more field drought experiments are urgently needed in the future, which would help us to reveal the physiological and ecological mechanisms of alpine vegetation in response to drought.

5. Conclusions

Quantifying the role of climate variables in shaping the vegetation response to drought is important for a comprehensive understanding of drought impacts. Using the NDVI and corresponding SPEI, we detected the spatial patterns of drought effects on alpine grasslands and analyzed their variations along the gradients of elevation and climate variables across different grassland types in the QTP. In general, drought imposed widespread lagged and cumulative effects on alpine grasslands. Meanwhile, the stronger sensitivity and short-term response time of vegetation were primarily distributed in the southwestern and northeastern parts of the plateau. Further, drought effects showed evident elevation dependence across different grassland types, which could be explained by altitudinal shifts in climate factors, including temperature and precipitation. Precipitation dominated the response of alpine meadows to drought, whereas temperature was the dominant driving factor of the vegetation response to drought on the alpine steppe. Our findings highlighted the divergent effects of hydrothermal conditions on different alpine grasslands in response to drought, which further implied a divergent change when facing future climate change. Additionally, more drought-related experiments are urgently needed to reveal the physiological and ecological mechanisms of alpine vegetation under drought.