Identifying the Responses of Vegetation Gross Primary Productivity and Water Use Efficiency to Climate Change under Different Aridity Gradients across China

Abstract

Despite the fact that gross primary productivity (GPP) and water use efficiency (WUE) have been widely used as indicators to evaluate the water-carbon cycle, uncertainties exist in the patterns of GPP and WUE responses to climate variability along different aridity gradients. In this study, the aridity index was used to divide China into four arid-humid zones. The spatiotemporal variability of multiple vegetation types GPP and WUE in response to climate change under different arid-humid zones were investigated based on remote sensing data. The results indicated that the increasing trend of WUE in the four arid-humid zones of China was less pronounced than GPP from 2001 to 2021. The GPP value decreased gradually from the humid to the arid zone, and the WUE value in the arid zone was slightly higher than in the semi-arid zone. The GPP of all vegetation types in China showed a tendency to increase, while shrubland and wetland WUE tended to decrease. The major vegetation types (e.g., forest, cropland and grassland) in each aridity gradient contributed to the changes in local GPP and WUE. However, in individual arid-humid zones, wetland and shrubland also exhibited high GPP and WUE values that were not inferior to forest and cropland. Temperature and precipitation were the main climatic factors responsible for the increase in vegetation GPP in different aridity gradients, with a higher positive correlation for temperature than precipitation. WUE showed a distinct positive and negative correlation with the thermal factors (temperature and net radiation) and the moisture factors (precipitation and relative humidity); this pattern was more pronounced in the humid and semi-humid zones. Net radiation and precipitation may be the main climatic factors causing a slight upward trend in WUE across the arid-humid zones, while the decrease in shrubland and wetland WUE may be related to relative humidity and precipitation.

1. Introduction

Since the 1860s, global warming exacerbated by massive carbon emissions has induced a series of variations in biogeophysical-chemical processes such as surface water balance, vegetation dynamics, and the coupled water-carbon cycle [1,2]. Studies have shown that the terrestrial biosphere has absorbed nearly one third of fossil fuel carbon emissions since 2000, which is an important component in mitigating the global carbon balance [3]. Furthermore, the response of vegetation in the biosphere to climate has been recognized as a central issue in global terrestrial ecosystem change [4]. The total amount of carbon fixed by vegetation through photosynthesis, which synthesizes carbon dioxide (CO₂) and water into organic compounds, can be replaced by gross primary productivity (GPP) [5]. Although we know that CO₂ fertilization can promote GPP, the GPP trend and influencing factors are still to be determined based on the frequency and intensity of external climate factors [6]. Meanwhile, vegetation carbon sequestration also means water loss, and water use efficiency (WUE) is an important indicator to quantify the coupling relationship between carbon sequestration and water loss [7]. Therefore, understanding the spatiotemporal variation of GPP and WUE as well as their responses to climate change will provide a reference for ecosystem management in the context of global warming.

WUE is usually considered to be the ratio of GPP to evapotranspiration (ET), and its accuracy is still limited by the data of GPP and ET [8]. Many studies on the spatiotemporal patterns of GPP and ET have summarized their measurement methods, which can be roughly classified into three types: in-situ measurements, remote sensing techniques, and model simulations [9]. The in-situ measurements are too limited to represent the overall ecological structure of the region, while the complex physical mechanisms and numerous parameters in the model simulations bring greater difficulty and uncertainty to the research work [10,11]. Remote sensing technology provides a reliable tool for large-scale GPP and ET assessment with the advantages of real-time transmission, rapid processing, and low-ground influence [12]. MOD16 is a global terrestrial evapotranspiration data product developed by the Penman-Monteith equation, with remotely sensed surface features. The accuracy rate of this dataset was 86%, as evaluated using the global flux measured data [13], which is advantageous for studying the spatiotemporal characteristics of ET on a regional scale. In addition, the MOD17 product of MODIS provides GPP data that have been widely used to study the global carbon cycle [14,15]. Liu et al. [16] compared this dataset with data from Ecosystem Model-Data Intercomparison (EMDI, obtained from the Oak Ridge National Laboratories) and 21 eddy flux measurement sites worldwide separately, yielding a high degree of consistency.

China added 1.35 × 10⁶ km² of new leaf area from 2000 to 2017, constituting 25% of the world’s net growth rate [17]. Meanwhile, with continued global warming, the northern boundary of the eastern subtropical and temperate zones in China has generally shifted northward, with profound effects on vegetation phenology [18]. Climate and vegetation changes in China can directly or indirectly alter its carbon sequestration and water cycle characteristics in terrestrial ecosystems [19]. Given the challenges of effective ecosystem management and addressing global change, a large number of studies have investigated the spatiotemporal variability of China’s vegetation GPP and WUE in recent years, and their responses to climatic factors. For example, Yao et al. [20] found that GPP in China increased significantly during 1982–2015 and the national GPP response to temperature was much higher than precipitation and radiation factors. Zhu et al. [21] concluded that GPP variations in China’s 128 vegetated ecosystems were different and highly sensitive to both precipitation and temperature. Regarding WUE, Sun et al. [22] divided China into eight different vegetation ecosystem zones and found an overall decreasing trend in WUE in China since 1982. Meanwhile, temperature had a greater influence on WUE across China, while specific humidity, precipitation, and short-wave radiation were only prominent in individual ecosystems. Li et al. [23] found that the WUE of China exhibited a slight upward trend since 2000, with precipitation influencing WUE to a greater extent than temperature. The above studies have established a foundation for understanding the state of GPP and WUE in China in a changing environment. However, due to the differences in spatiotemporal scales, climatic conditions and vegetation types, the interannual variation patterns of GPP or WUE and their responses to climatic factors varied among studies. Recent global examinations have reported the contradiction of GPP and WUE with aridity of different regions [24,25]. Liu et al. [16] also indicated that the responses of each vegetation ecosystem GPP and WUE to climate variables varied with the changes in hydrothermal conditions in different aridity gradients. Therefore, specific arid-humid zones can be delineated to further investigate the response of GPP and WUE to climatic factors for various vegetation types [26,27].

For observing GPP and WUE in specific spatial regions, studies are available that document the patterns of variation between the GPP and WUE across ecosystems, latitudes, or elevations [28,29]. It has been proposed that the frequency and intensity of aridity in northern, mid- and high latitudes will increase in the context of climate change. Drought can change both components (GPP and ET) of WUE, ultimately leading to its increase or decrease [30]. Studying the variation patterns of GPP and WUE in different drought gradients can lead to a better assessment of the response of the terrestrial ecosystem to climate change. Few studies in China have explored specific changes in GPP and WUE for various vegetation types in the context of different aridity gradients, and few have attempted to uncover the underlying mechanisms between climate change and these spatial variabilities [31]. There is a necessity to continue research in this important subject area. This not only enables a comprehensive exploration of spatiotemporal variability of GPP and WUE, but we also need to further our understanding of ecosystem responses to climate change as global drylands expand.

The main objective of this study is to investigate the spatiotemporal variability of GPP and WUE in response to climate change under different arid-humid zones. We intend to divide China into regions according to the aridity gradient, while identifying average temperature (Ta), Relative humidity (RH), precipitation (Pr), wind speed (WS), and net radiation (Rn) as the main climatic factors based on existing studies [16,26] to explore the response patterns of GPP and WUE to climate change in different arid-humid zones. The effects of land use/land cover change (LUCC) were not included in this study. In order to minimize the impact of LUCC, we extracted annual GPP and WUE values for each vegetation type based on land use/land cover data from the initial year (2001) [32]. The specific objectives were: (1) to ascertain the spatial and temporal variability of each vegetation type GPP and WUE in China’s different arid-humid zones during 2001–2021, and (2) to analyze the variation of climatic factors and their correlation with each vegetation type GPP and WUE in different arid-humid zones. This study can further elucidate the impact of climate change on GPP and WUE from the perspective of aridity gradient, which provides a reference for the management activities of ecosystems under climate change.

2. Materials and Methods

2.1. Study Region

China is a country with a vast and diverse physical geography, including a unique range of landforms, ecosystems and climate patterns. This country is low lying to the east, bordering the Pacific Ocean, and high lying to the west, deep in the hinterland of Eurasia. It has a land area of 9.6 million square kilometers, making it the third largest country in the world. The southeastern coastal regions of China are influenced by the southeast monsoon, which brings significant rainfall to these areas and is home to a range of tropical and sub-tropical monsoon climate ecosystems. In contrast, the northwestern part of China is controlled by a continental climate, resulting in dryness and a prevalence of low rainfall [33]. China’s unique climatic pattern has created a distinct southeast to northwest gradient of dryness and wetness, and has contributed to China’s unique natural ecosystems. As shown in Figure 1, the vegetation types such as forest are mainly distributed in the humid zone of China. Most cropland is distributed in the semi-humid zone, and there is extensive grassland in the semi-arid zone. Relatively little vegetation is found in the arid zone.

2.2. Data Sources

The Moderate Resolution Imaging Spectroradiometer (MODIS) products (https://modis.gsfc.nasa.gov/, accessed on 21 November 2022) analyzed in this study include GPP, ET, surface albedo, and land use/land cover data. WUE was calculated from the ratio of GPP to ET, where GPP and ET were obtained from MOD17A2H and MOD16A2 data for 2001–2021 with a temporal and spatial resolution of 8 days and 500 m. MOD17A2H data were produced by the Numerical Terra Dynamic Simulation Group, and a detailed description of the data algorithm can be found in previous studies [34]. The MOD16A2 algorithm was developed and improved based on the Penman-Monteith equation proposed by Mu et al. [35]. These products have been updated through a series of modifications and are now widely used in water-carbon cycle studies.

Surface albedo data was obtained from MCD43A3 with 500 m spatial resolution and a daily temporal resolution. This data was used for the calculation of net radiation. Land use/land cover data was derived from the MCD12Q1 at a yearly temporal resolution with 500 m spatial resolution. For the land use/cover map of China, we used the International Geosphere-Biosphere Project (IGBP) global vegetation classification dataset version in MCD12Q1, which defines 17 land use/land cover types including eleven natural vegetation, three non-vegetated land and three anthropogenic altered classes. The IGBP vegetation classification was reclassified into five vegetation types (Figure 1) according to the vegetation distribution characteristics across China, which was used to analyze the changes of GPP and WUE for China’s multiple vegetation types in the four arid-humid zones.

The meteorological data underlying this study include specific humidity, water vapor pressure, downward long/short wave radiation, temperature, precipitation, wind speed and potential evapotranspiration. Of these, long/short wave radiation was used for the calculation of net radiation combined with surface albedo data [7]. Relative humidity can be calculated based on specific humidity, water vapor pressure and temperature [36]. Details of the calculation are given in Equations (A1) and (A2) in Appendix A. This dataset was obtained from daily 0.25° Global Land Data Assimilation System (GLDAS) version 2.1 (https://ldas.gsfc.nasa.gov/gldas/, accessed on 21 November 2022) for 2001–2021, which was resampled to 500 m resolution using bilinear interpolation [37] to match the spatial resolution of the MODIS product.

2.3. Methodology

2.3.1. Aridity Index

To investigate the variation of vegetation GPP and WUE along the aridity gradient, we selected the aridity index (AI) to classify China into four arid-humid zones [27]. AI is calculated by precipitation/potential evapotranspiration based on the standard form proposed by the United Nations Environment Programme (UNEP), and is widely used in the classification of arid-humid zones and climate types. The four arid-humid zones across China were the humid zone (AI > 0.65), the semi-humid zone (0.5 < AI < 0.65), the semi-arid zone (0.2 < AI < 0.5), and the arid zone (AI < 0.2), as shown in Figure 1. The area proportion of all vegetation types in arid-humid zones is presented in Table S1.

2.3.2. Trend Analysis

The methods applied to test the trend changes in the time series of the factors can be divided into parametric methods, which require the test data to be independent and follow a normal distribution, and non-parametric methods, which only require the data to be independent [38]. Here, the Mann–Kendall test and Sen’s slope estimation method (two non-parametric methods) were used to analyze the variations of GPP, WUE and each climatic factor.

The Mann–Kendall test is used for trend analysis with the statistical value $Z_{MK}$ [39,40]. Positive and negative $Z_{MK}$ correspond to upward and downward trends. At a specific significance level $\alpha$, $\left|Z_{MK}\right|>Z_{1-\alpha /2}$ indicates a significant time series trend of the factor. In this study, the significance level $\alpha$ was set to 0.05, and the value of $Z_{1-\alpha /2}$ corresponding to this significance level was 1.96, corresponding to a confidence level of 95%. A confidence level of above 95% implies a significant trend of variation, otherwise it is not significant.

Sen [41] developed a non-parametric procedure that has been widely used along with the Mann–Kendall test to estimate the variation rate of relevant elements.

$$Q_{\mathrm{i}}=\frac{x_j-x_k}{j-k}$$

(1)

where $x_j$ and $x_k$ are the data values for sample $i$ ($j>k$); $i=1,2,\dots ,N$; $N=\frac{\mathrm{n}(n-1)}{2}$ where $n$ is the sequential year (21a), the $N$ values of $Q_i$ are arranged from smallest to largest, and the Sen’s slope is calculated as shown in Equation (2), where a positive or negative $Q_{med}$ value reflects an upward or downward trend.

$$Q_{med=}\{\begin{array}{ll}{Q}_{[(N+1)/2]},\hfill & ifNisodd\\ \frac{Q_{[N/2]}+Q_{[(N+2)/2]}}{2},\hfill & ifNiseven\end{array}$$

(2)

2.3.3. Coefficient of Variation

The coefficient of variation ($CV$) is an indicator to describe the fluctuation of data, which can reflect the stability of vegetation GPP and WUE over time.

$$CV=\frac{\sqrt{\frac{{\displaystyle \sum _{i=1}^n\left(x_i-\overline{x}\right)}}{n-1}}}{\overline{x}}$$

(3)

where $CV$ represents the coefficient of variation of GPP or WUE, $x_i$ represents the value of GPP or WUE in year $i$, and $\overline{x}$ represents the average value of GPP or WUE from 2001 to 2021.

2.3.4. Correlation Analysis

The Pearson correlation coefficient applied in this study to analyze the spatial distribution of correlations between GPP, WUE and each climatic factor in the selected period (2001–2021).

$$r_{xy}=\frac{n\times {\displaystyle \sum _{i=1}^n\left(x_i\times y_i\right)-{\displaystyle \sum _{i=1}^n{x}_i{\displaystyle \sum _{i=1}^n{y}_i}}}}{\sqrt{n\times {\displaystyle \sum _{i=1}^n{x}_i^2-{\left({\displaystyle \sum _{i=1}^n{x}_i}\right)}^2}}\sqrt{n\times {\displaystyle \sum _{i=1}^n{y}_i^2-{\left({\displaystyle \sum _{i=1}^n{y}_i}\right)}^2}}}$$

(4)

where $r_{xy}$ is the correlation coefficient between GPP or WUE and climatic factors, and $y_i$ represents the relevant climatic factors in year $i$.

3. Results

3.1. Spatiotemporal Variations of GPP and WUE

Figure 2 presents the spatial distribution of the multi-year average GPP and WUE across China from 2001 to 2021. Both exhibited manifest patterns of spatial variation (Figure 2a,d) and were in good agreement with the division of each arid-humid zone. Specifically, the average GPP and WUE across China from 2001–2021 were 692.65 gC⋅m⁻² and 1.17 gC⋅m⁻²⋅mm⁻¹, respectively. GPP decreased gradually from the humid zone to the arid zone (Figure 2b), with the humid zone (1083.75 gC⋅m⁻²) being 5.6 times higher than the arid zone (192.20 gC⋅m⁻²). The distribution pattern of WUE was similar to that of GPP, but the value in the arid zone (0.84 gC⋅m⁻²⋅mm⁻¹) is slightly higher than the semi-arid zone (0.78 gC⋅m⁻²⋅mm⁻¹) (Figure 2e), and the WUE in the humid zone (1.43 gC⋅m⁻²⋅mm⁻¹) was about twice as high as that in the arid and semi-arid zones.

Among all vegetation types, forest and cropland had higher GPP and WUE values of 1172.16, 756.95 gC⋅m⁻² and 1.52, 1.43 gC⋅m⁻²⋅mm⁻¹, respectively (Figure 2c,f). These two vegetation types occupied most areas of the humid zone (68.44%) and the semi-humid zone (39.43%), which were the main contributors to the GPP and WUE values in these two zones (Table S1). In contrast, grassland occupied most of the area in semi-arid regions (73.20%), but it had relatively small GPP (299.30 gC⋅m⁻²) and WUE (0.81 gC⋅m⁻²⋅mm⁻¹). The shrubland had the lowest GPP (291.25 gC⋅m⁻²) and WUE (0.62 gC⋅m⁻²⋅mm⁻¹) values among all vegetation types in China, while wetland GPP (327.64 gC⋅m⁻²) and WUE (0.64 gC⋅m⁻²⋅mm⁻¹) values were in the middle level. In general, the distribution of GPP and WUE across China had a strong geographical zonality along the aridity gradient, while, specifically within each vegetation type, they showed a more complex distribution of values.

A statistical analysis of GPP across China within the study years revealed a remarkable upward trend in GPP from 2001 to 2021 (6.41 gC⋅m⁻²⋅yr⁻¹). In comparison, the national WUE increased at a lower rate, with an average variation value of 0.0016 gC⋅m⁻²⋅mm⁻¹⋅yr⁻¹ (Figure 3 and Figure S1). Figure 4 and Figure 5 provide detailed statistics on the temporal changes of GPP and WUE in each arid-humid zone and vegetation type from 2001 to 2021. Regarding the variation of GPP, a significant upward trend was observed in all arid-humid zones and vegetation types (Figure 4a and Figure 5a). In each arid-humid zone, the greatest upward trend of GPP was displayed in the humid and semi-humid zones, with 7.63 and 7.56 gC⋅m⁻²⋅yr⁻¹, respectively. Arid and semi-arid zones showed a lower increase in GPP than humid and semi-humid zones (Figure 3). The tendency of increasing GPP was greatest in forest and cropland (7.91 and 8.71 gC⋅m⁻²⋅yr⁻¹) among all vegetation types, with the increasing trend in cropland being significantly greater than that in forest. From the area proportion of GPP in increasing regions around the country (Figure 4b and Figure 5b), the area proportions of increasing GPP in each arid-humid zone and vegetation type all reached more than 90% except for wetland, and the significantly increased regions (p < 0.05) also remained an area proportion of more than 50%.

Similar to GPP, WUE also exhibited an increasing trend in all arid-humid zones (Figure 4c), but showed a slightly increased value due to the smaller area proportion of WUE increasing regions in each aridity gradient compared to GPP (Figure 4d), which corresponded to the spatial trend of Figure 3c. Among the vegetation types (Figure 5c), WUE of forest, cropland and grassland showed increasing trends (0.0013, 0.0028, and 0.0016 gC⋅m⁻²⋅mm⁻¹⋅yr⁻¹, respectively). The three vegetation types occupied most of the area in the humid, semi-humid and semi-arid zones, which caused the rising values of WUE in these three regions (0.0015, 0.0022, and 0.0014 gC⋅m⁻²⋅mm⁻¹⋅yr⁻¹, respectively), to be extremely consistent with these three vegetation types. The area proportion of shrubland and wetland WUE decreasing regions was higher than increasing regions, which led to a national decreasing trend in their WUE (Figure 5c,d).

The $CV$ of GPP and WUE gradually increased from the humid zone to the arid zone (Figure 4b,d). Judging from the spatial distribution of GPP $CV$ (Figure 3b), although GPP increased significantly at the national level, the $CV$ of GPP was generally low (0.13), which indicated that there was no large fluctuation in GPP during 2001–2021. The regions with relatively high $CV$ in GPP were mainly concentrated in the central-eastern part of the semi-arid zone, which may be related to the implementation of national ecological restoration projects. The spatial distribution of WUE $CV$ in Figure 3d showed that the regions with large WUE fluctuation were mainly located in the southwestern part of the semi-humid zones of China. The average $CV$ of WUE on the national scale was lower (0.11) than GPP. Among different vegetation types, wetland exhibited higher $CV$ values for GPP and WUE (Figure 5b,d).

3.2. Changes in GPP and WUE for Each Vegetation Type in Arid-Humid Zones

To further investigate the variation of GPP and WUE along the aridity gradient for all vegetation types, we conducted statistical analyses on the average GPP and WUE values, trends, and $CV$ for all vegetation types in four arid-humid zones from 2001 to 2021, as shown in Figure 6 and Figure 7, and

Table 1. Trends (units: gC⋅m⁻²⋅yr⁻¹) and $CV$ of GPP for each vegetation type in different arid-humid zones from 2001 to 2021.

$\mathbf{GPP}-\mathbf{Trend}/\mathit{C}\mathit{V}$	Humid	Semi-Humid	Semi-Arid	Arid
Forest	7.94/0.103	7.69/0.111	5.65/0.096	4.15/0.158
Shrubland	1.62/0.251	14.78/0.126	0.49/0.112	1.97/0.221
Grassland	3.53/0.153	5.36/0.151	4.35/0.163	2.84/0.179
Wetland	1.35/0.182	3.89/0.219	4.54/0.108	3.69/0.124
Cropland	8.42/0.102	9.31/0.121	7.21/0.123	5.06/0.122

and

Table 2. Trends (units: gC⋅m⁻²⋅mm⁻¹⋅yr⁻¹) and $CV$ of WUE for each vegetation type in different arid-humid zones from 2001 to 2021.

$\mathbf{WUE}-\mathbf{Trend}/\mathit{C}\mathit{V}$	Humid	Semi-Humid	Semi-Arid	Arid
Forest	0.001/0.091	0.002/0.088	0.001/0.087	−0.001/0.161
Shrubland	−0.002/0.142	−0.001/0.081	0.001/0.136	−0.001/0.202
Grassland	0.002/0.155	0.001/0.132	0.002/0.132	0.002/0.136
Wetland	−0.001/0.199	0.006/0.203	−0.003/0.109	−0.002/0.103
Cropland	0.002/0.092	0.003/0.086	0.001/0.086	0.001/0.094

. In each of the arid-humid zones, GPP and WUE values remained high for forest and cropland, and maintained a stable distribution pattern for grassland, while wetland and shrubland varied dramatically. In the semi-humid zone, the values of GPP and WUE in the shrubland were significantly higher than other vegetation types and also exhibited a higher distribution compared to the other three arid-humid zones (Figure 6b and Figure 7b). In the arid and semi-arid zones, the GPP and WUE of wetland also showed the same distribution pattern. But this phenomenon is not evident in Figure 2 due to the relatively small area of both on a national scale.

Table 1 reveals that the GPP of all vegetation types in each arid-humid zones exhibited an increasing trend, among which shrubland in the semi-humid zone showed the largest GPP growth values, and wetland GPP also showed relatively high growth values in the arid and semi-arid zones. In all arid-humid zones, the GPP of forest and cropland generally increased significantly, with the GPP of cropland increasing to a greater extent in all but the arid zone, while forest was concentrated in the humid and semi-humid zones. Similarly, the grassland GPP maintained a steady increase across the arid-humid zones. As can be seen from Table 2, forest, cropland and grassland WUE in the different dry-wet zones mainly showed a slightly increasing trend, while shrubland and wetland mostly showed a decreasing trend, corresponding to the phenomenon of changing WUE of each vegetation in Figure 5c. The $CV$ of GPP for all vegetation was stable at 0.13. The $CV$ of wetland GPP and WUE were larger than those of other vegetation types in the semi-humid zone. Combining Figure 3b,d showed that GPP and WUE exhibited larger $CV$ values mainly in the southwestern part of this region, where wetland was more distributed (Figure 1), it can be further inferred that the large fluctuation of GPP and WUE in this region may be caused by wetland.

3.3. Climatic Effects on GPP and WUE

Figure 8 gives the spatial trends of five major climatic factors in China from 2001 to 2021. RH, Rn, and WS mainly showed a decreasing trend nationwide, while Ta and Pr displayed a manifest increasing trend. Among the five climatic factors, RH varied from 2.03% to 0.93%, with a decreasing trend in approximately 90% of the regions except for central and northeastern parts of the semi-humid and semi-arid zones. Rn tended to decrease in 66% of the country, mainly in the humid, semi-humid, and semi-arid zones of central-eastern China. The variation of Ta across China from 2001 to 2021 ranged from -0.05 °C to 0.32 °C. Ta exhibited an upward trend throughout 97% of China, with significant increases in all regions except the central and northeastern semi-humid and semi-arid zones. Pr and WS varied from −32.05 to 48.71 mm and −0.16 to 0.14 m·s⁻¹, respectively. Pr showed an increasing trend in 75% of the country. WS decreased in 73% of China, with a significant decrease in a larger proportion (55%). Details of the area proportion of change in each climatic factor during 2001–2021 can be found in Table S2.

Figure 9 and Figure 10 represent the spatial distribution of the correlation coefficients for GPP and WUE with each climatic factor. At the national scale, the average correlation coefficients of RH, Rn, Ta, Pr, and WS with GPP were −0.18, −0.06, 0.33, 0.19, and −0.16, respectively. Ta had the largest correlation coefficient with GPP and the trend was very consistent, as seen in the change curves of GPP and Ta over the period 2001–2021 (Figure S1). In all arid-humid zones and vegetation types (Figure 11 and Figure 12), GPP showed a high positive correlation with temperature. Although GPP also performed a higher sensitivity to RH, it was mainly concentrated in humid and semi-humid zones and exhibited a negative correlation with RH compared to Ta (Figure 9 and Figure 11). Besides Ta, GPP also positively correlated with Pr in all arid-humid zones and vegetation types, inferring that these were the two main climatic factors responsible for the upward trend of GPP from 2001 to 2021, but Ta contributed more to GPP than Pr. As can be seen in Figure 9 and Figure 11, the correlation coefficient between GPP and Pr exceeded that of Ta only in the semi-arid zone. Moreover, GPP showed high correlation with Ta for all vegetation types except grassland. Of the other two climate factors, GPP was negatively correlated with Rn and WS in most arid-humid zones and vegetation types (Figure 11 and Figure 12). The average correlation coefficient of GPP with WS was close to RH at the national level, but was not prominent in each dry and wet zone.

The average correlation coefficients of RH, Rn, Ta, Pr, and WS with WUE were −0.13, 0.14, 0.16, −0.10, and −0.02, respectively, at the national level. Figure 10 demonstrates a strong positive spatial correlation between WUE and the two thermal factors Rn and Ta, while a negative correlation with the moisture factors RH and Pr, a phenomenon that is more pronounced in the humid and semi-humid zones (Figure 11). Among the five climatic factors, the correlation coefficient between WUE and Ta was the largest, followed by Rn. Although the values were different, it can be seen from Figure 11 that the correlation coefficient of WUE with Ta and Rn were also the largest in each arid-humid zone and all showed positive correlations. While the negative correlations between WUE and RH, Pr, and WS were concentrated in the humid and semi-humid zones. The relationships between forest, grassland and cropland, which occupy most of the area in China, and the five climate factors also exhibited a consistent pattern (Figure 12). However, shrubland WUE showed similar magnitude of correlation coefficients with the five climatic factors, where it showed a higher positive correlation with RH. Wetland WUE had higher negative correlation values with Pr.

Figure 13 and Figure 14 further show the average correlation coefficients of GPP and WUE corresponding to each vegetation type with climatic factors in the four arid-humid zones. The specific values are presented in Tables S3 and S4. Similarly, we could find that the GPP for all vegetation in each arid-humid zone showed an obvious uniform positive and negative correlation with Ta and RH (Figure 13a,c). Rn and GPP exhibited a negative correlation in all arid-humid zones except for the humid zone, with the shrubland showing the most pronounced (Figure 13b). Pr was positively correlated with GPP at a higher level in relatively arid regions (Figure 13d). In Figure 14, except for shrubland and wetland, forest, grassland and cropland WUE showed positive correlations to Rn and Ta, with negative correlations to RH and Pr. This phenomenon was more obvious in the humid and semi-humid zones, while in the arid and semi-arid zones, the response of WUE to climate factors showed spatial heterogeneity among vegetation types. It is noteworthy that in the correlation analysis between WUE and Rn, all vegetation types in each arid-humid zone exhibited positive correlations (Figure 14b). Wetland WUE showed higher negative correlation with Pr and RH (Figure 14a,d), which coincided with the higher average correlation coefficient of WUE with Pr and RH for wetland in Figure 12. In summary, the GPP and WUE of most vegetation types in all arid-humid zones across China were sensitive to Ta. GPP and WUE also showed high sensitivity to Pr and Rn, respectively. The responses of some vegetation GPP and WUE to a climatic factor varied in different regions, which might be related to local topography, vegetation density and the growth cycle.

4. Discussion

Exploring the spatiotemporal variability of vegetation GPP and WUE responses to climate change in different arid-humid zones across China can further reveal the patterns of water-carbon cycle in each vegetation ecosystem, help predict how ecosystems will respond to future climate scenarios, and provide a reference for conservation and management strategies [42]. MODIS products with continuous time series and high resolution offer the potential for large-scale GPP and WUE studies. MODIS GPP and ET datasets have been updated with a series of modifications to capture the spatiotemporal patterns of GPP and WUE for different vegetation types and climatic conditions, which have been widely used in regional or global water-carbon cycle studies.

4.1. Distribution and Dynamics of GPP and WUE

This study found that the spatial distribution of GPP and WUE across China was in good agreement with the delineation of each arid-humid zone, suggesting that the aridity gradient has a significant impact on GPP and WUE across China (Figure 2). GPP decreased steadily from the humid zone to the arid zones, while the WUE was slightly higher in the arid zone than the semi-arid zone. In general, the humid zone is well-watered, allowing plants to perform adequate photosynthesis, resulting in higher GPP values. Conversely, the arid zone receives less precipitation and tends to have lower soil moisture levels, which may limit plant growth and reduce GPP. This phenomenon has been well documented in previous studies [43]. In addition, Yu et al. [44] indicated a slight improvement in vegetation WUE under drought stress, which can lead to a higher WUE in the arid zone than the semi-arid zone. The main vegetation types in each arid-humid zone contributed most of the GPP and WUE values (Figure 2, Figure 6 and Figure 7); forest with a larger leaf area, higher canopy structure, and deeper root distribution had higher GPP and WUE than other vegetation types [45]. China’s cropland is vast and dominated by irrigated production, contributing to relatively high GPP and WUE [46]. The GPP and WUE of forest and cropland were prominent in different arid-humid zones, but shrubland also showed higher values in semi-humid zones, which was similar to the study of Sun et al. [22]. This study inferred that the semi-humid zone, which lies between the arid and humid zones, usually has a moderate amount of rainfall, which is not sufficient to support the growth of tall plant species. This means that shrubs adapted to these conditions can be more efficient than forest in using water and nutrients to maintain high levels of GPP and WUE [47]. In addition, shrubland in the semi-humid zone usually has small, thick leaves and long root whiskers, which allow shrubland to maintain high levels of photosynthesis while minimizing water loss, further increasing its GPP and WUE [48]. Furthermore, in semi-arid and arid zones, the wetland tended to have stronger GPP and WUE than the common substrate. Wetland in arid zones is often characterized by low water availability, and wetland plants and microorganisms may have adapted to use water more efficiently. This may result in higher photosynthetic rates as well as higher water use efficiency [49]. Despite their small individual size, grassland had a large distribution in each arid-humid zone, which caused their overall average GPP to be similar to that of shrubland and wetland, while WUE was higher than that of shrubland and wetland.

GPP and WUE exhibited an upward trend from 2001 to 2021 across the country as well as in each arid-humid zone, but it is also found that China’s ET showed an upward trend during this period [12], which may be the direct reason for the weaker rise in WUE (Figure 3, Figure 4 and Figure 5). We summarized the ET variation patterns in recent years across China to further investigate the mechanisms of GPP and WUE changes. It was found that the central, southwestern and northeastern parts of the semi-humid and semi-arid zones have been the main regions of increased ET in China since the 21st century [50], with a similar spatial distribution to the GPP significant increase regions obtained in this study. Among these regions, WUE also exhibited increasing trends to different degrees. We inferred that implementing ecological restoration projects and related agricultural production modification in these regions led to an increase in GPP, WUE and ET [51,52]. In contrast, the regions with a high distribution of cropland in the humid zone (e.g., the Sichuan basin), GPP and ET showed an increasing trend while WUE mainly exhibited a decreasing trend, indicating that the increasing trend of GPP since 2001 was less than that of ET. A similar pattern was found in previous studies [23,53], which may be related to irrational agricultural water use in these regions. Specifically for the five main vegetation types in China, all vegetation types exhibited an increasing trend in GPP, while wetland and shrubland WUE showed a decreasing trend. Studies have shown that a warmer climate and a longer growing season can stimulate photosynthesis and plant growth to increase vegetation GPP, yet this also increases plant transpiration and thus reduces WUE [54].

An analysis of GPP and WUE coefficients of variation since 2001 revealed that both showed large $CV$ values in the southwestern part of the semi-humid zone (Figure 3). Since wetland was more distributed in this region (Figure 1) and its $CV$ of GPP and WUE was prominent among all vegetation types (Table 1 and Table 2), it can be inferred that the variation of wetland was the main reason for the significant fluctuation of GPP and WUE in this region. Most studies consider forests and agricultural land as the main vegetation types affecting the variations in GPP and WUE across China [7]. This study investigated the GPP and WUE dynamics of multiple vegetation types in arid-humid zones across China and found that wetland and shrubland also have their unique characteristics in specific arid-humid zones, which can be used to further decipher the mechanisms of China’s GPP and WUE changes.

4.2. Climatic Controls on GPP and WUE

Ta and Pr were the main climatic factors contributing to the increase in vegetation GPP in each aridity gradient. Ta maintained an increasing trend under all aridity gradients across China (Figure 8c), and GPP also kept a high positive correlation with Ta across arid-humid zones and vegetation types (Figure 13c). This study concluded a stronger positive correlation between Ta and GPP than Pr, which differs from Beer et al. [55]. Zeng et al. [56] recorded that the sensitivity of GPP to precipitation decreased in recent years with the gradual increase in rainfall. As can be seen in Figure 9d and Figure 11, this phenomenon was more pronounced in the humid zone where Pr was higher, and the positive correlation between GPP and Pr gradually increased as the humid zone transitioned to the arid zone. Furthermore, Figure 13d also revealed that wetland GPP showed a negative correlation with Pr in humid and semi-humid zones. Research has shown that excessive Pr can limit nutrient uptake by reducing the oxygen supply to the soil. In wetland ecosystems, prolonged inundation can cause oxygen stress on plant roots, leading to a decrease in GPP [57]. RH is the ratio of actual humidity to saturation humidity. In general, as the temperature rises, the air can hold more moisture and the RH decreases [58]. A negative correlation between RH and GPP can be found in Figure 13a for each arid-humid zone and vegetation type. This can reflect the contribution of rising Ta to GPP from the perspective of RH. Rn also exhibited a decreasing trend in the country (Figure 8b). Lower Rn could affect the photosynthesis of vegetation and consequently its GPP [59]. But the promotion of GPP by other factors (e.g., Ta and Pr) caused a negative correlation between Rn and GPP in the arid-humid zones where Rn mostly showed a decrease (Figure 13b). In addition, high WS can increase transpiration and respiration rates of plants, resulting in more water loss through leaves and offsetting some of the carbon gained through photosynthesis. WS has mostly shown a decreasing trend in all arid-humid zones since 2001, resulting in a negative correlation between GPP and WS for multiple vegetation types in each aridity gradient.

The correlation pattern between WUE and the five climatic factors was obvious in the aridity gradient, with the highest correlation coefficients with the two thermal factors Ta and Rn, followed by the moisture factors Pr and RH, and the lowest with WS. (Figure 11). Among all vegetation types, except wetland and shrubland, all other vegetation types also showed a high correlation with thermal factors (Figure 12). WUE was positively correlated with the thermal factor and negatively correlated with the moisture factor. This phenomenon gradually diminished from the humid zone to the arid zone (Figure 14). There is a critical value for the effect of Ta on WUE, and either too high or too low Ta is detrimental to plant WUE [60]. As shown in Figure 8c, the intensity of Ta increase in China is higher in arid and semi-arid zones than in humid and semi-humid zones. The study inferred that excessive warming led to inconsistent positive and negative correlations of WUE to Ta for each vegetation type in arid and semi-arid zones, while a uniform positive correlation was observed in humid and semi-humid zones (Figure 14c). Rn maintained a consistent positive correlation with WUE across all aridity gradients. Since 2001, Rn has shown a decreasing trend in the humid, semi-humid and semi-arid zones of east-central China (Figure 8), which may be the main climate reason for the decrease in WUE in this region (Figure 3). Excessive precipitation is also detrimental to the WUE of vegetation. In recent years, increased Pr in China has led to a negative correlation between Pr and WUE in several arid-humid zones (Figure 11). The highest negative correlation coefficient between WUE and Pr was found in wetland among vegetation types (Figure 12 and Figure 14d), and Pr may be the main climatic factor responsible for the decreasing trend of wetland WUE. Combined with the fact that the wetland was more distributed in the semi-humid zone of southwest China (Figure 1), and that it exhibited large $CV$ for WUE (Table 2), we inferred that the large fluctuation of WUE in this region was caused by the effect of Pr on wetland WUE. The correlation distribution of individual vegetation WUE and RH was more consistent with Pr across arid-humid zones (Figure 14a,d). Unlike Pr, RH mostly showed a decreasing trend in China (Figure 8a). Lower RH will promote the outward transpiration of water from plants to reduce WUE. This is different from the conclusion obtained in this study that WUE showed a negative correlation with RH in each arid-humid zone, which may also be related to the contribution of other factors (e.g., Ta) to WUE. Shrubland WUE displayed a positive correlation with RH (Figure 12 and Figure 14a). Although the correlation coefficients between shrubland WUE and the five climatic factors were similar in magnitude, it was relatively high with RH, and presumably the decline in shrubland WUE since 2001 may be related to the decrease in RH.

It should be noted that the correlation analysis, which can directly reflect the correlation between the two factors, was selected for this study. However, the relationship between each climatic factor on WUE as well as GPP is complex, and the five climatic factors selected in this study are not completely independent; their interactions may also play a role in the variations of GPP and WUE [61]. Future research should consider this issue to better determine the impact of climate change on GPP and WUE. Most previous studies have focused on the correlation of GPP and WUE with Ta and Pr [16,62]. Although this study also found a high correlation of GPP and WUE with Ta and Pr, the main climatic factors affecting GPP and WUE were somewhat different in each arid-humid zone and vegetation type. Besides the major vegetation types, there are some regional characteristics of GPP and WUE values of wetland and shrubland in response to climatic factors in different aridity gradients. Therefore, the response of GPP and WUE to climatic factors in each vegetation type can be further determined from the perspective of aridity gradient.

4.3. Uncertainties and Limitations

Based on MODIS and GLDAS data, this study analyzed the dynamics of GPP and WUE in different aridity gradients and the responses to climate change across China from 2001 to 2021. Although data such as MODIS have been validated to meet the needs of most studies, the use of remote sensing data inevitably introduces some potential uncertainties and limits the results of final spatiotemporal analyses [63]. At the same time, we have analyzed the variations of GPP and WUE in China for only 21 years since 2001 owing to the limited available data, and the understanding to support both responses to climate change is still inadequate. Furthermore, since the main objective of this study was to investigate the effects of climate change on GPP and WUE in different arid-humid zones, we extracted annual GPP and WUE values for each vegetation type using land use/land cover data from the initial year (2001) in order to exclude the effects of LUCC as much as possible [64,65]. In real situations, however, it is difficult to completely separate the effects of climate change and LUCC on vegetation GPP and WUE; LUCC inevitably has an impact on the optimization and validation of the results [66]. In the next study, the interactions of LUCC-related processes with vegetation GPP and WUE under climate regulation need more research.

5. Conclusions

This paper explored the spatiotemporal variation of GPP and WUE in China from 2001 to 2021. China was divided into four arid-humid zones based on the aridity index, and the effects of climate change on GPP and WUE in different arid-humid zones and multiple vegetation types therein were analyzed. In our investigations, we found the following:

(1)

From 2001 to 2021, the increasing trend of WUE in the four arid-humid zones of China was less pronounced than GPP. The GPP value gradually decreased from the humid to arid zone, while WUE in the arid zone was slightly higher than that in semi-arid zone due to drought stress. The $CV$ of GPP and WUE increased gradually from the humid zone to the arid zone, and the relatively high $CV$ values of both were mainly concentrated in the central-eastern part of the semi-arid zone and the southwestern part of the humid and semi-humid zones.

(2)

In all arid-humid zones, forest and cropland possess higher values of GPP and WUE. For individual zones and vegetation, shrubland in the semi-humid zone and wetland in the arid and semi-arid zone showed higher GPP and WUE values, which may be strongly related to the regional environment as well as to the structure of the vegetation itself. Across China, all arid-humid zones and vegetation types exhibited an upward trend for GPP, while shrubland and wetland WUE showed a downward trend.

(3)

Among the five climatic factors, Ta and Pr trended upward from 2001 to 2021, while RH, Rn, and WS trended downward in most parts of China. Ta and Pr were the main climatic factors responsible for the increase in GPP for each vegetation type in different aridity gradients. Ta had a stronger positive correlation than Pr, but the positive correlation of Pr gradually increased with the transition to the arid zone. WUE showed obvious positive and negative correlations with thermal factors (Ta and Rn) and moisture factors (Pr and RH), this pattern being more pronounced in humid and semi-humid zones. The decrease in Rn and increase in Pr may be the main climatic factors contributing to the weaker upward trend in WUE across the arid-humid zones, while the decrease in shrubland and wetland WUE may be related to RH and Pr. Overall, climate change may affect GPP and WUE differently along the aridity gradient, causing cross-gradient differences among ecosystems.