Asymmetric warming has different impacts on vegetation growth across various vegetation types on the Tibetan Plateau

ABSTRACT

Global warming could affect vegetation growth, while land surface temperature has exhibited an asymmetric warming pattern over the past 50 years. The Tibetan Plateau, known as “the roof of the world,” experiences the surface warming almost twice as high as the global average. However, previous research has largely overlooked the impacts of this asymmetrical warming on vegetation growth and the temporal changes in these impacts. In this study, we assess the effects of asymmetric warming on vegetation growth at regional and different vegetation types by using partial correlation analysis, and reveal the temporal changes of impacts strength using time moving window. The results showed that there had been a significant greening trend (1.01 × 10⁻³ yr⁻¹, p < 0.01) in the Normalized Difference Vegetation Index (NDVI) during the growing season, as well as significant warming trends in both maximum temperatures (T_max, 0.0354°C yr⁻¹, p < 0.01) and minimum temperatures (T_min, 0.0333°C yr⁻¹, p < 0.01) during 2000–2021. Under the background of asymmetric warming, the grassland showed stronger impacts of T_min than T_max on NDVI, while the opposite effects were observed for forests. Over time, the response of NDVI to T_max and T_min has intensified. This study highlights importance of asymmetric warming in the projection of climate change on vegetation in similarly vulnerable ecosystems worldwide.

KEYWORDS:

• Alpine vegetation
• asymmetric warming
• vegetation types
• Tibetan Plateau

1. Introduction

Vegetation plays a pivotal role in terrestrial ecosystems, exerting significant influence over global and regional carbon cycles and energy balances (Forzieri et al. 2020; Ge et al. 2021; Piao et al. 2019). There are significant changes in vegetation cover due to the influences of climate change and human activities (Liu et al. 2023; Zhu et al. 2022). Over the past three decades, the trend of increasing vegetation greenness, referred to as “greening” has been observed (Chen et al. 2019; Ding et al. 2020; Piao et al. 2019). Model simulations suggest that the 70% of the global greening could be attributed to the fertilization of carbon dioxide (CO₂) (Zhu et al. 2016). However, for high latitudes and the Tibetan Plateau, climate change remains a critical factor of vegetation greening, accounting for more than 28% of the global vegetation greening.

The global climate have showed warming trends in recent decades by using observation and reanalysis air temperature data (You et al. 2021; Zhang, Qin, and Zhai 2023) and remote sensing land surface temperature data (Sobrino, Julien, and García-Monteiro 2020; Y.-R. Wang et al. 2022), as well as the rate of warming on the Tibetan Plateau exceeds twice the global average (You et al. 2021). Widespread climate warming affected the vegetation growth on the Tibetan Plateau (Li et al. 2018; Wu et al. 2023; Xie et al. 2024; Zhang, Yang, and Long 2023). During 1982–2018, growing season temperature exerted the largest contribution to Normalized Difference Vegetation Index (NDVI) across the Tibetan Plateau (Li et al. 2022). Additionally, multiple vegetation indices revealed that the vegetation growth was dominated by temperature during 2000–2021 on the plateau (Wu et al. 2023). The eastern grassland is primarily limited by temperature, whereas the climatic constraint in the western is predominantly water availability the Tibetan Plateau (Xie et al. 2024). However, current research on vegetation growth often emphasizes the role of mean temperature, overlooking the influence of asymmetric warming on vegetation growth.

The recent study finds a reversal of the global daytime and nighttime warming pattern (Zhong et al. 2023). Furthermore, it was revealed that asymmetric warming showed different effects on vegetation growth and exhibited spatial heterogeneity (Du et al. 2019; Liu et al. 2023; Shen et al. 2016; Xu et al. 2023). For instance, in most moist and cool ecosystems of the Northern Hemisphere, NDVI demonstrated a positive correlation with daytime air temperature (the maximum daily temperature), and a negative correlation with nighttime air temperature (the minimum daily temperature) (Peng et al. 2013). Conversely, nighttime warming showed a positive effect on ecosystem productivity of grassland ecosystems in northern China based on air temperature (Du et al. 2019). On the Tibetan Plateau, nighttime warming has been found to positively affect summer NDVI in grasslands (Li et al. 2023; Shen et al. 2016). These studies focused on the effect of asymmetric warming on grasslands or regional scale. The Tibetan Plateau contains diverse vegetation types, ranging from broadleaf forests, to vast grasslands, and ultimately transitioning to deserts northward. However, the response of vegetation growth to asymmetric warming on the Tibetan Plateau among these varied vegetation types remains unclear. Furthermore, considering that growing season contributes significantly to vegetation growth and carbon sequestration (Liu et al. 2023), it is imperative to assess the impacts of asymmetric warming during the critical season and evaluate the differences among vegetation types.

Furthermore, many studies on the response of vegetation activity to asymmetric warming temperature have been stayed on static analysis (Xia et al. 2018; Xu et al. 2023). Indeed, the relationship between vegetation growth and temperature change is dynamic, evolving over time as a result of other environmental and human interventions. Research indicated a diminishing correlation in vegetation activity and temperature (Piao et al. 2014). Specially, the contribution of daytime temperatures was greater than that of nighttime temperatures in the weakening relationship between warming and vegetation growth in the temperate zone of the north of 30°N (Zheng et al. 2021b). Moreover, there are the different responses of temperature on vegetation activities in alpine meadows and steppes of the Tibetan Plateau as time changes (Cong et al. 2017). Therefore, changes in the vegetation growth responding to asymmetric warming over time need to be further investigated.

Above these studies, there are still gaps in understanding of how asymmetric warming impacts vegetation growth on the Tibetan Plateau and how these relationships between warming and vegetation growth change over time. Consequently, this study is centered around answering two questions: (1) whether the impact of asymmetric warming on vegetation growth varies among different vegetation types; (2) whether the intensity of asymmetric warming on the vegetation growth has changed over time.

2. Materials and methods

2.1. Study area

Tibetan Plateau is located in southwestern China and is known as “the third pole,” with an overall average altitude of over 4,000 meters and is the unique ecological units in the world (Yao et al. 2022). The vegetation types are complex and diverse. The natural vegetation is predominantly comprised by alpine grasslands (alpine steppes and meadows), which account for 50% of the plateau’s total area (Figure 1). The horizontal belt of the plateau is characterized by forests, scrubs, grasslands and deserts spanning from southeast to northwest, and the vertical belt has changed from humid type in the south edge of Himalayas to the continental arid type in the interior of the plateau.

Figure 1. The vegetation types on the Tibetan Plateau. The blank region, identified by multiyear mean annual NDVI values below 0.10, was emblematic of sparse vegetation. The vegetation types was derived from vegetation map of the People’s Republic of China (1:1000000) (X. Zhang 2007).

2.2. Framework

In this study, we first used the MODIS NDVI, EVI and reanalysis climate datasets to investigate the spatial and temporal variation of vegetation growth and climate factors during the growing season (May to September). To address the challenge posed by the high correlation and inter-feedback among climate factors, we employed the ridge regression method to determine the main driving factors of vegetation greenness. Given the dominance of temperature as a driving factor, we further explored the impact of daytime and nighttime temperature on vegetation growth using partial correlation analysis, which aims to examine the relationship between two variables and eliminate the influence of other variables. Furthermore, we assessed the impact difference of daytime and nighttime temperature on vegetation greenness among the different vegetation types. Finally, we investigated the changes in vegetation greenness in response to daytime and nighttime temperature at different period (Figure 2).

Figure 2. Framework of the study.

2.3. Datasets

2.3.1. Vegetation greenness

The NDVI and EVI were quantified as vegetation greenness, which were sourced from MOD13A1 version 6.1, available at a spatial and temporal resolution of 500 m and 16 days (Didan 2021). We used the maximum value composition method to calculate the monthly NDVI to mitigate the effects of clouds and shadows (Cao et al. 2022). In this study, we excluded the pixels with mean annual NDVI values below 0.10 due to their sparse vegetation (Zhang et al. 2021). The growing season (May–September) mean NDVI as a proxy of vegetation growth (Liu et al. 2023). The EVI was calculated in the same ways.

2.3.2. Climate data

The climate data, including air temperature, precipitation, and solar radiation downward from 2000 to 2021, was extracted from the ERA5-Land dataset, with a temporal resolution of 1 hour and a spatial resolution of 0.1°. The ERA5-Land dataset has better performance in capturing inter-annual variability of temperature and precipitation (Jiao et al. 2021; Lv et al. 2024; Y. Peng et al. 2023). Then, we firstly calculated the maximum daily temperature (daytime temperature, T_max) and the minimum daily temperature (nighttime temperature, T_min) data by the hourly climate data, and then the monthly temperature was averaged by daily temperature per month from 2000 to 2021. The monthly precipitation and solar radiation downward data were calculated by cumulative way based on hourly data. Furthermore, we calculated the growing season T_max and T_min by mean ways during May to September per year, and growing season cumulative precipitation (Pre) and growing season cumulative solar radiation downward (Ssrd) were calculated by cumulative ways during May–September per year. Then, all the dataset were matched with the NDVI dataset using geographical coordinates.

2.3.3. Other data

The vegetation types data was derived from Vegetation Map of the People’s Republic of China (1:1000000) (Zhang 2007). The contiguous solar-induced fluorescence (CSIF) dataset was obtained from Zhang et al. (2018) and then we calculated the growing season CSIF during 2000–2019 (Lv et al. 2023; Zhang et al. 2023). The soil moisture dataset with 0~5 cm depth was derived from Shangguan et al. (2023) during 2001–2020, with a temporal resolution of 1 day and a spatial resolution of 1 km, matching with the NDVI dataset by using geographical coordinates and then we calculated the multiyear mean of soil moisture.

2.4. Analyses

2.4.1. Temporal changes in vegetation greenness and climate factors

For each pixel, the temporal trend of growing season NDVI from 2000 to 2021 was assessed by ordinary least squares regression (Peng et al. 2012; Zheng et al. 2021). To analyze the temporal change of climate factors, we also calculated the trends of growing season T_max, T_min, precipitation and solar radiation downward from 2000 to 2021 by the same ways.

2.4.2. Determine the dominant climate factors on vegetation greenness

To overcome the challenge posed by the strong correlation and mutual feedback among various climate factors, ridge regression is a robust method for parameter estimation for solving the covariance problem in multiple linear regression (Chen et al. 2020). This method enhances the reliability of regression coefficients by eliminating the biases inherent in the least squares method (Graham 2003). We used ridge regression to examine dominant climate factors between growing season NDVI and T_max, T_min, precipitation, and solar radiation downward. Furthermore, the dominant factor was identified based on the variable whose absolute value of the standardized regression coefficient was the highest.

2.4.3. Relationship between daytime and nighttime temperature on vegetation growth

Partial correlation analysis is a statistical method used to explore the relationship between two variables (such as variables X and Y) while considering the influence of the third variable (such as variables Z) on them. The formula as:

(1) $\mathit{r}=\frac{{\mathit{r}}_{\mathit{XY}}-r_{\mathit{XZ}}{r}_{\mathit{YZ}}}{\sqrt{1-{r^2}_{\mathit{XZ}}}\sqrt[\mathit{\hspace{0.17em}}]{1-}{r^2}_{\mathit{YZ}}}$(1)

In the formula, r is the Partial correlation coefficient between variables X and Y and $r_{\mathit{XY}}$, $r_{\mathit{XZ}}$, $r_{\mathit{YZ}}$ are the simple correlation coefficient between variables X and Y, X and Z, Y and Z.

Therefore, we used partial correlation analysis to evaluate the impact of the T_max of growing season (May to September) on NDVI controlling the T_min, total precipitation and solar radiation downward of growing season (Shen et al. 2016). The impact of the T_min on growing NDVI was calculated by the same ways. All the variables were detrended (Peng et al. 2013). The sensitivity of vegetation greenness to growing season T_max and T_min was quantified as its coefficient in a multiple regression analysis that assessed the impacts of growing season T_max, T_min, precipitation and solar radiation downward on NDVI. NDVI gradually reaches saturation and provides a poor estimate of vegetation parameters with high vegetation cover. Even if the ground vegetation continues to grow, the NDVI does not increase with the change or only changes slightly (Haboudane et al. 2004), while the EVI continued to exhibit sensitivity to canopy variations (Huete et al. 2002). We also test the robustness of results by using EVI datasets. We used the time moving window assess the change of impacts of T_max and T_min on NDVI during 2000–2021.

3. Results

3.1. Spatial-temporal variations in vegetation greenness and climate factors

Most of study area experienced increasing trends in vegetation greenness during growing season (Figure 3). We found that NDVI increased in 75.2% of all pixels during 2000–2021, particular in northeast, center and east, while a decreasing trend was observed in 24.8% of all pixels, mainly in the southwest and south (Figure 3a). In 30.1% and 2.2% of all pixels, NDVI showed a significantly increasing and decreasing trend (p < 0.05). In total, the growing season NDVI significantly increased by 1.01 × 10⁻³ yr⁻¹ during 2000–2021 (p < 0.01, Figure 3b). We found similar results when we used EVI, NIRv, LAI and SIF datasets (Lv et al. 2023).

Figure 3. (a) Spatial distribution of trends in growing season NDVI (NDVI_GS) and (b) temporal variation in NDVI_GS on the Tibetan Plateau during 2000–2021.

The increasing trends in regional averaged T_min (0.0354°C yr⁻¹, p < 0.01) were larger than that of T_max (0.0333°C yr⁻¹, p < 0.01) in growing season during 2000–2021. The regional average of precipitation showed an increasing trend, while the average solar radiation downward showed a decreasing trend, but neither trend was statistically significant. Spatially, growing season T_max and T_min increased in 93.5% and 99.1% of all pixels widespread in Tibetan Plateau, of which 49.5% and 59.1% were significant (p < 0.05, Figure 4a,b). We further investigated the trends of growing season precipitation and solar radiation downward (Figure 4c,d). We observed 69.9% and 30.1% of all pixels displayed increasing and decreasing trends in growing season precipitation, of which 3.3% and 0.1% were significant at p < 0.05 level (Figure 4c). In contrast, the solar radiation downward showed decreasing trends in 88.2% of all pixels (12.7% significant).

Figure 4. Spatial distribution of trends in growing season (a) mean maximum temperature (T_maxGS), (b) minimum temperature (T_minGS), (c) cumulative precipitation (Pre_GS), and (d) cumulative solar radiation downward (Ssrd_GS) during 2000–2021. The inset was trend of T_maxGS, T_minGS, Pre_GS and Ssrd_GS at p < 0.05 level.

3.2. Response of daytime and nighttime temperature on the vegetation greenness

To quantify the proportion of the areas dominated by multiple drivers, we calculated the absolute value of the contribution of the explanatory variables by using ridge regression. The results showed that NDVI was dominated by T_max, T_min, precipitation and solar radiation downward in 28.7%, 28.8%, 19.9%, and 22.6% of the pixels (Fig. S1). The contributions of temperature (T_max + T_min) on NDVI were largest, exceeding half of the study area, mainly distributed in the southeast, center, east and some areas in west of the plateau. This implied that temperature dominates vegetation growth during the growing season. Due to the asymmetric warming, we further compared the partial correction between daytime and nighttime temperature and vegetation greenness.

During 2000–2021, for 42.2% of all pixels, the NDVI showed a positive correlation with T_max (2.8% significant at p < 0.05), distributing in the southeast and south of the Tibetan Plateau (Figure 5a). A negative correlation between NDVI and T_max was observed in 57.8% of all pixels (6.3% significant at p < 0.05), primarily concentrated in northwest, center and northeast regions. With regards to the T_min, NDVI exhibited a positive correlation with T_min in 58.8% of all pixels, with significance observed in 6.5% of the pixels, predominantly in northwest, center and northeast regions. The negative correlations between NDVI and T_min were located in southeast, and the negative correlations were significant in 2.5% of the pixels (Figure 5b). To quantify the proportion of the areas dominated by T_max and T_min, we calculated synergistic effect of the explanatory variables (Fig. S2). During 2000–2021, the dominant factor influencing vegetation dynamics was the negative correlation between NDVI and T_max alongside the positive correlation between NDVI and T_min (T_max-, T_min+) was the dominant factor for 50.2% of all pixels, distributed in northwestern and eastern of alpine grasslands. For 33.6% of all pixels, NDVI exhibited a positive correlation with T_max and a negative correlation with T_min (T_max+, T_min-), mainly in southeastern and southern regions. Additionally, in 8.7% of all pixels, NDVI was positively correlated with both T_max and T_min in some areas in central plateau. NDVI was negatively correlated with T_max and T_min in 7.6% of the pixels in northwestern regions. We also found the NDVI exhibited stronger sensitivity to T_min compared to T_max (Figure 5c,d).

Figure 5. Spatial distribution of (a) and (b) partial correlation coefficient between growing season NDVI and maximum temperature (T_max) and minimum temperature (T_min), (c) and (d) sensitivity of growing season NDVI to T_max and T_min. variation in the proportion of pixels on partial correlation between growing season NDVI and (e) T_max and (f) T_min in different vegetation types (>5% of pixels). (g) variation (median: line; mean: solid circle) in changes in the partial correlation coefficient and (h) sensitivity between growing season NDVI and T_max (the blue box) and T_min (the orange box) in different vegetation types (>5% of pixels).

The proportion of pixels with a positive correlation between NDVI and T_max was lowest in the steppes, followed by meadows, scrubs, needleleaf forests, and broadleaf forests. In contrast, more than 60% of the pixels in the steppe’s areas of exhibited a positive correlation between NDVI and T_min, followed by meadows and scrubs, and forests were the lowest proportion of positive correlation between NDVI and T_min. With the increase of vegetation cover, the proportion of pixels with a positive correlation between NDVI and T_max increased, while T_min showed a opposite distribution (Figure 5e,f).

Furthermore, the partial correlation coefficient and sensitivity of NDVI to T_min were larger than that to T_max among the entire region and different vegetation types, especially in alpine steppes and meadows regions, while, the partial correlation coefficient of NDVI to T_max was larger than that to T_min in the forest region (Figure 5g,h). These results suggested that T_min had a more positive effect on grasslands than forests. The same result was also exhibited by using EVI dataset (Fig. S3).

In addition, we assessed the relationship between CSIF and T_max and T_min. To match NDVI dataset, the CSIF pixels (with 0.05° spatial resolution) were reserved that corresponding all NDVI pixels (with 500 m spatial resolution) have values. During 2000–2019, the CSIF was positively correlated with T_min and negatively correlated with T_max in grasslands and there was a reverse relation in forest (Fig. S4). This spatial pattern was similar to NDVI. Furthermore, we also examine the effect of CO₂ to NDVI during 2000–2020 controlling with growing season T_max, T_min, precipitation and solar radiation downward. During 2000–2020, for 48.2% of all pixels, the NDVI was positively correlated with CO₂ (3.6% significant), and the NDVI was negatively correlated with CO₂ accounting for 51.8% of all pixels (4.1% significant, Fig. S5). The CO₂ does not always have a positive impact on vegetation and even has a negative impact in most areas.

3.3. Changes in vegetation greenness in response to daytime and nighttime temperature at different period

We further investigated the response of NDVI to T_max and T_min across different periods (Figure 6). We tested the changes of the partial correlation coefficients using moving window from 7 to 12 years. We identified the 9-year moving window as the most suitable for our analysis, given the significant changes observed in the partial correlation coefficients. We found that partial correlation coefficients between NDVI and T_max decreased (p < 0 05) at an average rate of 0.056 yr⁻¹ from 2000 to 2021 under the 9-year moving window (Figure 6a), and the partial correlation coefficients between NDVI and T_min increased at an average rate of 0.060 yr⁻¹ (p < 0.05, Figure 6b). Over the past 20 years, there were significantly increasing trends in response of NDVI to both T_max and T_min on the Tibetan Plateau, and the magnitude of changes of the partial correlation coefficient between NDVI and T_min was higher than that of T_max but stabilized in recent years. Spatially, the relationship between NDVI and T_max/T_min experienced increases, mainly concentrated in the grassland areas in the northeastern, east-central and northwestern of the Tibetan Plateau (Figs. S6 and S7).

Figure 6. (a) Temporal variations in partial correlation coefficient between growing season NDVI and maximum temperature and (b) minimum temperature from 7-year–12-year moving window.

4. Discussion

4.1. Asymmetric effects of temperature in response to vegetation greenness on the Tibetan Plateau

From 2000 to 2021, the growing season T_max and T_min showed significantly increasing trends (p < 0.05) on the Tibetan Plateau, and there was an asymmetry warming that the rate of warming at nighttime was higher than that daytime. The findings align with the previous studies (You et al. 2016). In the grassland of the Tibetan Plateau, there was a higher proportion of pixels with a positive correlation between NDVI and T_min compared to those with a positive correlation between NDVI and T_max during the growing season (Figure 5g,h). Additionally, the partial correlation coefficient of NDVI on T_min was observed to be higher than that on T_max (Figure 5e,f). These findings suggested that T_min had a greater impact on vegetation than T_max in the grassland region of the Tibetan Plateau. Conversely, in the forest region, the opposite result was observed, where T_max had a greater impact on vegetation than T_min.

First, the warming of nighttime temperatures (T_min) could reduce the risk of frost especially in grassland (Shen et al. 2016). The multiyear mean of the days of T_min <0°C is more than 210 days in more than 75% of the region of the Tibetan Plateau and the number of days of T_min <0°C increased from the southeast to the northwest (Figure 7a). Steppes have the highest number of T_min <0°C at 298 days, following by meadows. The broadleaf forests show the smallest number of T_min <0°C at 89 days (Figure 7c). During 2000–2021, the number of days with a minimum temperature below 0°C shows a decreasing trend in 96.9% of all pixels (56.8% significant, p < 0.05), widely distributed over most of the study area (Figure 7b). Fewer days below 0°C in grasslands than in forested areas (Figure 7d). Therefore, grassland areas are exposed to greater frost risk than forest areas. The nighttime warming reduces the frequency of low temperatures and protects vegetation against frost damage (S. Peng et al. 2013; Shen et al. 2016).

Figure 7. Spatial distribution of (a) multiyear mean of days of growing season mean minimum temperature (T_min) < 0°C and (b) trends of days of T_min < 0°C on the Tibetan Plateau during 2000–2021, (c) multiyear mean of days of T_min < 0°C and (d) trends of days of T_min < 0°C among different vegetation types (>5% of pixels).

Secondly, vegetation growth in grasslands was still limited by low temperature in the Tibetan Plateau. We observed a positive correlation between NDVI and T_min where T_min was below 0°C in grasslands (Figure 8a), and vegetation growth might be limited by low T_min. However, only 2.05% of forest areas have T_min <0°C and the correlation between NDVI and T_min is close to 0 (Figure 8b). At the same time, very low temperatures could damage the cell structure of the vegetation (Shen et al. 2016), therefore higher T_min could promote vegetation growth by reducing the limitation of low temperature in grasslands than forests.

Figure 8. Variations of partial correlation coefficient (R_P) between growing season NDVI and minimum temperature (T_min) with minimum temperature in (a) grasslands (including steppes and meadows) and (b) forests (including needleleaf forests, broadleaf forests).

In addition, nighttime warming could enhance leaf respiration, leading to greater consumption of synthesized carbohydrates. This phenomenon may trigger a compensatory stimulation of photosynthesis in the following day (S. Peng et al. 2013; Zheng et al. 2022). Further, nighttime warming has been shown to be strongly associated with increased humidity and precipitation (Cox et al. 2020; Liu et al. 2023). Global asymmetry study suggested that nighttime warming was often accompanied by wetter climatic conditions, possibly due to changes in cloud cover (Cox et al. 2020). Thus, nighttime warming may lead to enhanced moisture effectiveness and promote vegetation growth.

The negative effect of daytime temperature on NDVI in grassland areas may be affected by warming-induced drought stress. In this study, we found that grassland NDVI was positively correlated with T_max when aridity < 1.00, then the NDVI correlation between NDVI and T_max showed a negative value while the negative relation was stronger from −0.06 at aridity of 1.00 ~ 1.25 to −0.19 at aridity > 10.00 in drier areas (Figure 9a). Furthermore, NDVI was negatively correlated with T_max in low growing season precipitation lower than 750 mm and the partial correlation coefficient changed to positive value and remained stable with the increase of precipitation. The variations of relation between growing season NDVI and T_max with soil moisture showed same pattern as precipitation. This suggests that the aridity may be influenced by surface soil moisture. Warmer daytime temperatures enhance evapotranspiration from vegetation and reduce soil surface moisture (Peng et al. 2013), and at the same time, warmer daytime temperatures also lead to an increase in the water vapor pressure difference. The lack of water vapor pressure coupled with lower precipitation in the region may reduce the time of stomatal opening (Peng et al. 2013; Zheng et al. 2022), which may increase the drought stress and then affect the vegetation growth.

Figure 9. Variations of partial correlation coefficient between growing season NDVI and maximum temperature (T_max) with (a) aridity, (b) multiyear mean of growing season precipitation during 2000–2021 and (c) multiyear mean of growing season soil moisture during 2001–2020 (due to data gaps in 2000 and 2021) on the Tibetan Plateau grasslands (including steppes and meadows).

In forest areas, T_max has a greater effect on vegetation than T_min. This is likely attributed to the prevailing temperatures not reaching the optimal temperature for vegetation growth (Chen et al. 2021). Moreover, the regions have sufficient precipitation, so daytime warming can effectively increase photosynthetic enzyme activities, soil nitrogen effectiveness, and thus increasing the rate of plant photosynthesis and thus growth (Melillo et al. 2002; Turnbull, Murthy, and Griffin 2002).

4.2. Changes in the effect of asymmetric warming on vegetation greenness on the Tibetan Plateau

Based on the results of partial correlation and least squares regression using a 9-year moving window, it has been found that T_max has a negative impact on NDVI, whereas T_min has a positive impact on NDVI in the Tibetan Plateau from 2000–2021 (Figure 6). As time passes, there is a significant increasing trend for the response of NDVI to T_max and T_min of the growing season. It is worth noting that vegetation growth may be limited by low temperatures and frost risk (S. Peng et al. 2013; Shen et al. 2016). As nighttime temperatures increase, it becomes beneficial for vegetation to reduce temperature limitation and frost risk, which in turn promotes vegetation growth and leads to an increase in vegetation greenness. Consequently, this leads to an increase in the relation of NDVI to T_min. However, higher daytime temperatures, possibly due to excessive evapotranspiration, coupled with precipitation below 600 mm in most regions of the Tibetan Plateau (Wang et al. 2020), would increase drought stress and slow down vegetation growth. This would eventually lead to an increase in the absolute value of the negative correlation coefficient of NDVI to T_max. However, we also found the strength remained stable in recent years, which might be due to greater adaptability to temperature under warming conditions (Shi et al. 2015).

4.3. Uncertainties and prospects

In this study, we concentrated the impact of warming on vegetation growth using the ridge regression method to determine the main climate factors. However, the environmental and anthropogenic factors are complex. Vegetation growth is also affected by precipitation (Sun et al. 2019), nitrogen deposition (Yang et al. 2011), land use changes (Song, Feng, and Wang 2022) and human activities (Huang et al. 2022; Pan et al. 2017; Y. Wang et al. 2022; Yu et al. 2016). Since 2003, grazing separation by fencing has become a common practice on the Tibetan Plateau to grassland restoration. It has been reported that about one third of the northern Tibetan area was fenced between 2004 and 2012, and the fencing could increase the above-ground biomass of degraded grasslands (Y. Wang et al. 2022; Yu et al. 2016). In addition, reforestation have also increased the vegetation cover in the Tibetan Plateau (Song, Feng, and Wang 2022). In the future, we should reinforce the comprehensive evaluation of vegetation growth. Moreover, we also found the different response of asymmetric warming on vegetation growth between grasslands and forests. This could provide a foundation for improving performance of vegetation dynamic models.

5. Conclusions

In this study, we analyzed the vegetation greenness change of growing season response to daytime and nighttime warming during 2000–2021 by using ridge regression and partial correlation analysis method. Furthermore, we assessed the changes of relationship between vegetation greenness and asymmetric warming over time. The findings clearly indicate that vegetation greenness showed an increasing trend during 2000–2021 and is influenced by temperature in more than half of the Tibetan Plateau. The daytime and nighttime temperatures of growing season in the Tibetan Plateau showed a significantly increasing trend (p < 0.05), but the rate of nighttime warming was higher than that during the daytime. In the grassland region of the Tibetan Plateau, nighttime temperature has a stronger effect on vegetation growth than daytime temperatures, likely attribution to a reduction in low-temperature limit and frost risk by nighttime warming. The positive effect of daytime warming on vegetation limited by precipitation due to drought stress on grassland regions. In forest areas, higher daytime temperatures can promote photosynthesis in vegetation due to the abundance of precipitation. With global warming, there is a greater effect between daytime and nighttime temperature on vegetation.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All the data used for this study are publicly available online. MOD13A1 NDVI and EVI, https://lpdaac.usgs.gov/products/mod13a1v061/; ERA5-Land dataset, https://cds.climate.copernicus.eu/; Elevation data, https://lpdaac.usgs.gov/products/srtmgl1v003/; vegetation types data, https://www.plantplus.cn/dsite/zhibei/b12.html.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15481603.2024.2348257.

Additional information

Funding

The work was supported by the Second Tibetan Plateau Scientific Expedition and Research Program [2019QZKK0405].