The Contributions of Climate and Human Activities to Water Use Efficiency in China’s Drylands

Abstract

In the context of global warming, terrestrial ecosystems have undergone significant variations. China has implemented a variety of ecological engineering methods to enhance carbon stocks. However, understanding the spatial and temporal dynamics of carbon and water in drylands under climate change remains limited. Here, our research elucidates carbon and water dynamics in China’s drylands over the last two decades, with a focus on understanding spatial–temporal changes and the effects of ecological engineering on the carbon–water cycle. Furthermore, this study investigates the relationships among climate change, water use efficiency (WUE), and its components—Gross Primary Productivity (GPP) and Evapotranspiration (ET)—identifying key climatic drivers and assessing possible directions for enhancing WUE under changing climate conditions. Our research indicates that both GPP and ET have significantly increased over the past 20 years, with growth rates of 4.96 gC·m⁻²·yr⁻¹ and 4.26 mm·yr⁻¹, respectively. Meanwhile, WUE exhibited a slight declining trend, at a rate of −0.004 gC·mmH₂O·yr⁻¹. This confirms the positive impact of vegetation restoration efforts. We found that fluctuations in interannual WUE were influenced by human activities and climate change. Precipitation (Prec) was the key climatic factor driving the GPP increase. Both solar radiation (Solra) and Prec were crucial for the interannual variation of WUE. Interestingly, WUE was the main factor affecting GPP development. The decline in WUE in drylands is linked to interannual variability in WUE and increased Vapor Pressure Deficit (VPD) due to warming. Seasonal variations in how WUE responds to climatic factors were also observed. For instance, fall rainfall increased WUE, while spring rainfall decreased it. Fall WUE was highly sensitive to VPD. Spatially, we found higher WUE in China’s eastern and Xinjiang regions and lower in inland areas and the Tibetan Plateau. Geomorphologic factors and soil conditions were the main drivers of this spatial variability in WUE. Temperature (Tem), Solra, VPD, and relative humidity (Relah) also played significant roles. Our results show a generalized inverse persistence in WUE variability. This suggests a potential for increased WUE in the eastern regions and a risk of decreased WUE on the Tibetan Plateau. Addressing the threat of vegetation decline in arid regions, particularly within the Tibetan Plateau, is crucial. It is essential to adapt forestry practices to complement the carbon and water cycles in these landscapes.

1. Introduction

Plant water use efficiency (WUE) is a measure of the amount of carbon fixed per unit of water consumed [1,2]. At the ecosystem level, the WUE is typically calculated as the ratio of gross primary productivity (GPP) to evapotranspiration (ET) [3,4]. As a vital indicator, WUE elucidates terrestrial ecosystems’ adaptation to environmental shifts, spotlighting the carbon–water nexus [5,6]. In the context of global warming, drylands exhibit a more pronounced warming trend than humid regions [7]. Consequently, adverse effects of warming are anticipated to occur more frequently and intensely in drylands. Drylands, home to over a third of the global population and covering nearly half the land, rank among the most climate-vulnerable ecosystems [8]. Previous studies have revealed that variations in WUE in drylands are sensitive to climate change and human activities. Therefore, monitoring WUE dynamics is critical for unraveling dryland ecosystems’ carbon and water cycles under changing environmental conditions [9,10].

Natural factors and human actions significantly influence WUE [11,12]. Previous studies have demonstrated that WUE is closely related to the composition and diversity of plant communities but is more sensitive to climate change than plant structure [13,14,15]. Furthermore, the mechanisms by which dryland vegetation responds to various climatic factors exhibit significant heterogeneity across regions [7,16]. Precipitation (Prec) and temperature (Tem) stand out as primary influencers of WUE variability. For example, a study reported that WUE was closely related to Prec in temperate and tropical drylands [17], while the limiting effect of Tem on WUE was more pronounced in high-latitude regions. In addition, certain studies have revealed the existence of a crucial Tem threshold for WUE variability, beyond which increases in Tem result in decreased WUE [18]. The expansion of arid regions due to rising Tems contributes to greening of these areas, but it also exacerbates atmospheric desiccation. This affects changes in WUE in drylands. Under global warming, VPD plays a significant role in WUE by affecting ET. Low plant VPD affects WUE by altering plant transpiration and soil evaporation, whereas high VPD induces stomatal closure, thereby reducing plant photosynthesis in ecosystems [19,20], particularly in drylands. Solar radiation (Solra) plays a significant role in determining the capacity of ecosystems to sequester carbon by directly controlling vegetation photosynthesis, which can significantly affect WUE [21]. Some scholars have found that there was a threshold for the role of Solra in increasing WUE [18], beyond which WUE decreased with further increases in Solra. Human interventions, notably through land cover changes, profoundly affect WUE by altering ecosystems’ carbon sequestration and water dynamics [22]. On agricultural land, human activities, such as irrigation and fertilizer application, as well as an increase in crop yields, can result in alterations to the carbon and water cycles, which can increase WUE [23]. Simultaneously, overgrazing has a significantly negative impact on grassland ecosystems, particularly alpine meadows, causing grassland degradation and desertification [24,25]. Several studies have demonstrated that certain ecological measures such as reforestation initiatives in China have contributed to the greening trend in the northern regions [26]. This has also led to an increase in water consumption. Moreover, there were regional uncertainties in the direction of the change in WUE.

China’s drylands, spanning over 2.4 × 10⁶ km², face critical environmental challenges. However, extensive restoration efforts since 1978 have significantly mitigated, if not reversed, desertification [27]. Concurrently, large-scale afforestation, as a primary restoration measure, has resulted in considerable changes in WUE for dryland vegetation by affecting GPP and ET [28]. Furthermore, the climate in this region has exhibited a trend toward warming and wetting, which in turn has affected WUE dynamics [29]. Hence, studies examining the effects of climate change and human activities on WUE in drylands in China have garnered considerable attention. For instance, a study utilizing eddy covariance (EC) tower data and remote sensing-based Normalized Difference Vegetation Index (NDVI) data to investigate the spatial variation in WUE across aridity gradients suggested that Prec and elevation were the primary factors of this spatial variation [30]. Another study on the WUE dynamics and driving factors in the Tibetan Plateau in China’s drylands between 1981 and 2010 indicated that climate change was the primary factor behind changes in WUE, with Tem being the most significant factor [31]. Additionally, a report indicated that increases in NPP were largely affected by changes in Prec, and factors including Prec and Tem also dominated the changes in WUE [32]. Nevertheless, current studies have only focused on the effects of climate change and human activities on WUE in China’s drylands at the single scale, and few studies have comprehensively investigated the driving forces behind changes in WUE at time series and spatial differentiation scales. Moreover, the precise extent to which climate change and human activities have contributed to variations in WUE in China’s drylands remains unclear. Therefore, there is a pressing need to conduct extensive research on the complex factors that underlie changes in WUE.

This study thoroughly examined the effects of climate change and human interventions on Water Use Efficiency (WUE) across China’s arid regions from both spatial and temporal perspectives, quantifying the degree of their contributions. The primary objectives of this study were as follows: (1) to analyze the spatial and temporal patterns of WUE in China’s drylands using remote sensing and meteorological data; (2) to quantitatively evaluate the contributions of various climatic factors and human activities to the changes in WUE and to investigate the underlying mechanisms of multi-year spatial differentiation in WUE; and (3) to explore the effects of changes in WUE on China’s dryland ecosystems and to predict future ecosystem changes. The results of this research offer a solid scientific foundation for the conservation of ecosystems and strategies for adapting to climate change. They deepen our comprehension of how China’s dryland ecosystems are evolving in response to global warming and help to guide the execution of ecological initiatives within the country.

2. Materials and Method

2.1. Study Area

China’s drylands mainly include the northeastern, northwestern, and southwestern regions. Drylands in China cover 14 provinces, occupying an area of 6.525 × 10³ (Figure 1). Vegetation types of this region are primarily grassland (40.13%), desert vegetation (22.28%), cultivated vegetation (12.52%), and alpine vegetation (5.05%) (Figure 1a). Most of the area belongs to the middle temperate belt, south temperate belt, and plateau climate zone (Figure 1b), of which the climate types are primary temperate continental and temperate monsoon climates. The landforms are diverse, with altitude differences exceeding 8000 m (Figure 1c). From 2001 to 2020, the annual average Tem was between 18 °C and 22 °C. The annual average Prec varied from 0 to 1000 mm, and VPD ranged from 0 to 4.76 kPa. The variation in VPD spanned 0 to 4.76 kPa. China’s drylands are categorized into hyper-arid (AI ≤ 0.05), arid (0.05 < AI ≤ 0.2), semi-arid (0.2 < AI ≤ 0.5), and semi-humid (0.5 < AI ≤ 0.65) subtypes (Figure 1c) [33,34]. Human activities have significantly impacted the ecological environment of China’s drylands. Numerous ecological engineering measures have been taken to enhance the region’s fragile ecology. Afforestation plays a crucial role in mitigating land degradation. The expansion of forested areas is directly linked to afforestation efforts and the conversion of farmland back to forest [29]. At the same time, under the stimulation of economic development, the inevitable expansion of towns and cities will change the nature of the subsurface and thus affect the ecological environment.

2.2. Data and Pre-Processing

The datasets for GPP, ET, NDVI, and LAI were obtained through Google Earth Engine (GEE). GPP and ET datasets underwent yearly summation processing within the GEE to generate annual GPP and ET datasets with a spatial resolution of 1 km. After maximum synthesis on the GEE platform, NDVI and LAI data were resampled to a 1 km spatial resolution. Meteorological data were collected daily from 402 stations within the study area (Figure 1a). This dataset selected variables that included air Tem (i.e., max, min, and mean Tem, Prec, sunshine duration, average relative humidity (Relah), and wind speed). The AI index was defined as the ratio of potential evapotranspiration (PET) (mm) to Prec [33]. The Penman–Monteith equation was utilized to compute daily PET. Daily VPD was determined by calculating the difference between saturation and actual vapor pressures. Spatial interpolation of the aforementioned data was conducted using ANUSPLIN 4.2 software. Meteorological raster maps from 2001 to 2020 were then obtained, with a common spatial resolution with WUE data of 1 km.

A soil water content (SWC) (m³·m⁻³) dataset was derived from the fifth-generation product of the European Centre for Medium-Range Weather Forecasts (ECMWF) Global Climate Atmospheric Reanalysis (ERA5), with data originating from the global mid-term weather forecasts. It is the volume fraction of soil moisture. It offers volumetric fractions for soil water content at four depths: 0–7 cm, 7–28 cm, 28–100 cm, and 100–289 cm, at a 0.1° spatial resolution [35]. Additionally, the monthly annual average SWC was calculated, and the dataset for 2001 to 2020 was obtained. In China’s drylands, soil water within a 100 cm depth serves as the primary source for plant growth [36,37,38]. Therefore, this study utilized the average soil water content for depths of 0–100 cm to investigate the effect of SWC on changes in WUE. Finally, the SWC dataset was resampled to a common spatial resolution of 1 km. Slope and aspect were extracted from elevation (Elev) data. The datasets for Elev, slope, and aspect were resampled to a 1 km resolution. Land Cover type (Lanct) datasets were download from the Zenodo website. Vegetation type (Veget), soil type (Soilt), geomorphic type (Geomt), gross domestic product (GDP), and population (POP) data were all from Recourse and Environment Science and Data Center.

Table 1. Data sources.

Type	Source	Product	Resolution	Time Series
Meteorological data	China Meteorological Data Service Center (https://data.cma.cn/)	V3 Datest	1 day	2001–2020
GPP	Google Earth Engine (GEE) (https://earthengine.google.com)	NASA LP DAAC MOD17A2H	500 m 8 days	2001–2020
ET	Google Earth Engine (GEE) (https://earthengine.google.com)	NASA LP DAAC MOD16A2	500 m 8 days	2001–2020
LAI	Google Earth Engine (GEE) (https://earthengine.google.com)	NASA LP DAAC MOD15A2H	500 m 8 days	2001–2020
SWC	European Centre for Medium-Range Weather Forecasts (ECMWF) (https://www.ecmwf.int/)	Global Climate Atmospheric Reanalysis (ERA5)	0.1° 1 month	2001–2020
Land Cover (Lanct)	Zenodo (https://zenodo.org/)	CLCD (Yang and Huang, 2021)	30 m 1 year	2001–2020
Vegetation type (Veget)	https://www.resdc.cn/		1 km	2001 (Publication time)
Soil type (Soilt)	https://www.resdc.cn/		1 km	1995
Geomorphic type (Geomt)	https://www.resdc.cn/		1 km	2009 (Publication time)
Elevation data (Elev)	Geospatial Data Cloud (https://www.gscloud.cn)	SRTMDEMUTM 90M	90 m	—
Gross Domestic Product (GDP)	Recourse and Environment Science and Date Center (https://www.resdc.cn/)	Kilometer Grid Dataset of China’s GDP Spatial Distribution	1 km	2000, 2005, 2010, 2015, 2019
Population (POP)	Recourse and Environment Science and Date Center (https://www.resdc.cn/)	Kilometer Gridded Dataset of Chinese Population Spatial Distribution	1 km	2000, 2005, 2010, 2015, 2019

provided more details on the specific sources of these datasets. All the data utilized in this study shared a uniform spatial resolution of 1 km and were projected onto the same spatial projection system.

2.3. Method

2.3.1. Calculation of WUE

This study introduces WUE to explore how vegetation productivity interacts with total water consumption in China’s drylands. A commonly used expression for WUE is the ratio of GPP (gC·m⁻²) to ET (mm) [4]:

$$WUE=\frac{GPP}{ET}$$

(1)

2.3.2. Temporal and Spatial Trend Analysis of WUE, GPP, and ET

In order to elucidate the temporal and spatial variations in WUE, GPP, and ET within the drylands of China, this study employed Theil–Sen (Sen) trend analysis and the Mann–Kendall (MK) test. These methodologies exhibit robust capabilities in mitigating the impact of outliers and measurement errors. The formula for calculating the slope is as follows:

$$slope=\mathrm{Median}\left[\frac{x_j-x_i}{j-i}\right],1\le i\le j\le n$$

(2)

Median is the median function. n is the number of years. There is a total of 20 years from 2001 to 2020; i and j are time series numbers; $x_i$ and $x_j$ are the values corresponding to the variable x in the i-th and j-th years respectively. A slope greater than 0 indicates an increasing trend in WUE, whereas a slope less than 0 signifies a decreasing trend. The significance of trends is evaluated at a 95% confidence level, corresponding to a significance level where the $\left|Z\right|$ statistic is greater than 1.96. Based on the outcomes of trend analysis and significance testing, the patterns of WUE variation in this study were categorized into four distinct modes: significantly increasing (slope > 0, $\left|Z\right|$ > 1.96), insignificantly increasing (slope > 0, $\left|Z\right|$ < 1.96), significantly decreasing (slope < 0, $\left|Z\right|$ > 1.96), and insignificantly decreasing (slope < 0, $\left|Z\right|$ < 1.96).

Especially in water-scarce regions, there exists a complex equilibrium relationship among WUE, GPP, and ET. Vegetation degradation leads to reduced GPP. Initially, plants may close stomata, decreasing ET and increasing WUE. Later, increased surface bareness reduces WUE [39]. In the later stages of degradation, increased surface bareness results in reduced WUE. As vegetation recovers and the ecosystem stabilizes, ET decreases [40]. To further elucidate the mechanisms behind WUE variations, separate Sen’s slope trend analyses were conducted for GPP and ET. The changing mechanisms of WUE within China’s drylands were classified into six distinct patterns (

Table 2. The classification of WUE variation patterns.

Pattern Name	Trend
Pattern A	${slope}_{WUE}<$$0; {slope}_{GPP}<$$0; {slope}_{ET}<$0	GPP dominant reduced WUE
Pattern B	${slope}_{WUE}<$$0; {slope}_{GPP}<$$0; {slope}_{ET}>$0	GPP, ET reduced WUE
Pattern C	${slope}_{WUE}<$$0; {slope}_{GPP}>$$0; {slope}_{ET}>$0	ET dominant reduced WUE
Pattern D	${slope}_{WUE}>$$0; {slope}_{GPP}<$$0; {slope}_{ET}<$0	ET dominant increased WUE
Pattern E	${slope}_{WUE}>$$0; {slope}_{GPP}>$$0; {slope}_{ET}<$0	GPP, ET increased WUE
Pattern F	${slope}_{WUE}>$$0; {slope}_{GPP}>$$0; {slope}_{ET}>$0	GPP dominant increased WUE

Note: ${slope}_{WUE}$, ${slope}_{GPP}$, and ${slope}_{ET}$ represent the slopes obtained from Sen’s trend analysis for WUE, GPP, and ET, respectively.

).

2.3.3. The Quantitative Relationships between Climatic Factors and WUE, GPP, and ET

To explore the response of WUE, GPP, and ET to climate change, this study utilized partial correlation analysis to examine the degree of association between them and climatic factors. Partial correlation analysis was employed to investigate the relationship between two variables, x and y, while simultaneously controlling for the influence of other factors on variable x [27]. This study employed partial correlation analysis to determine the correlation between WUE and four climatic factors: Tem, Prec, Solra, and VPD. The formula for calculating the partial correlation coefficient was as follows:

$$r_{xy.z_1{z}_{2\dots }{z}_g}=\frac{r_{xy.z_1{z}_{2\dots }{z}_g}-r_{x{z}_g.z_1{z}_{2\dots }{z}_{g-1}}·r_{y{z}_g.z_1{z}_{2\dots }{z}_{g-1}}}{\sqrt{\left(1-{r_{x{z}_g.z_1{z}_{2\dots }{z}_{g-1}}}^2\right)·}\sqrt{\left(1-{r_{y{z}_g.z_1{z}_{2\dots }{z}_{g-1}}}^2\right)}}$$

(3)

$r_{xy.z_1{z}_{2\dots }{z}_g}$ represents the partial correlation coefficient between variables x and y after eliminating the influence of variables z₁, z₂, …, zg. A partial correlation coefficient greater than 0 suggests a positive correlation between WUE and the climate factor, while less than 0 suggests a negative correlation.

In order to quantitatively describe the influence of climate factors on WUE, GPP, and ET, this study calculated the contribution rate of each climate factor to the change of WUE, GPP, and ET using the following method [41,42]:

$$\mathrm{Contr}.{\mathrm{Y}}_{\mathrm{j}}=\frac{|R_{y_j{x}_i}|}{\sum _{i=1}^{i=n}|R_{y_j{x}_i}|}\times 100\%$$

(4)

Y_j is the contribution rate of climate factors to the dependent variables, including WUE, GPP, and ET. $R_{y_j{x}_i}$ is the partial correlation coefficient between climate factors and the dependent variables. Four climate factors, Prec, Tem, Solra and VPD, were considered in our study.

2.3.4. Multiple Linear Regression and Residual Analysis

In this study, residual trend analysis was employed to distinguish the effects of climate variability and human activities on WUE. This method, proposed by Evans et al. [43], has been widely applied and is based on the principle of establishing a model between vegetation WUE and climatic factors. The impact of human activities on vegetation is characterized by a variance that cannot be explained by the model.

In past residual analyses, the selected meteorological factors were mostly focused on Tem, Prec, and Solra. Given the significant controlling effect of VPD on WUE, this paper adds VPD to the residual analysis on top of the three meteorological factors. This analysis helps quantify the impact of climate change and human activities on WUE’s temporal variations in China’s drylands, elucidating WUE’s temporal drivers. The relevant calculation formula is as follows:

$${WUE}_{cli}\left(i,t\right)=\mathrm{a}\times Tem\left(i,t\right)+\mathrm{b}\times Pre\left(i,t\right)+\mathrm{c}\times Solar\left(i,t\right)+\mathrm{d}\times VPD\left(i,t\right)+e$$

(5)

$${WUE}_{res}\left(i,t\right)={WUE}_{obs}\left(i,t\right)-{WUE}_{cli}\left(i,t\right)$$

(6)

where ${WUE}_{cli}\left(i,t\right)$ is WUE predicted value by regression of climate factors, representing WUE value under climate change, and Tem, Pre, Solra, and VPD are four climate factors. ${WUE}_{obs}\left(i,t\right)$ is the observed value of WUE; ${WUE}_{res}\left(i,t\right)$ is the residual value, defined as the WUE value under the influence of human activities.

Next, the slopes of observed value, predicted value, and residual value were recorded as SlopeWUEobs, SlopeWUEcli, and SlopeWUEres. The three factors were used to calculate the contribution rate of climate change and human activities to WUE in China’s drylands. The calculation method is shown in

Table 3. Approach for determining the proportional impact of climate variability and human actions on WUE alterations.

Driving Force Partition			Contribution of Driving Forces/%
SlopeWUEobs	SlopeWUEcli	SlopeWUEres	Climate Change	Human Activities
>0	>0	>0	SlopeNPPcli/Slope NPPobs	Slopewueres/slopewueobs
	>0	<0	100	0
	<0	>0	0	100
<0	<0	<0	−slopewueobs/slopewueobs	−Slopewueres/slopewueobs
	<0	>0	−100	0
	>0	<0	0	−100

Note: slope_NPPobs represents the slope of the true value of WUE; slope_NPPcli represents the slope of simulated values of WUE under the control climatic factors; Slope_NPPres represents the slope of WUE residuals values, i.e., the slope in WUE under the control of human activities.

.

2.3.5. Geographical Detector

Various factors, including topography, climate, and land cover changes, subject WUE to multiple influences. The interactions among these factors further compound their impact [26]. This study employed a geographic detector to quantify the impacts of individual factors and unravel the driving forces behind the spatial variations in WUE. Specifically, 21 factors were selected to investigate the spatial driving forces of WUE within China’s drylands. These factors were Tem, Prec, Solra, PET, VPD, SWC, ET, GPP, Relah, AI, LAI, NDVI, Soilt, Veget, Elev, Slope, Aspect, Geomt, Lanct, GDP, and POP. The factor detector measured the contribution level of each factor to the spatial diversity in WUE, shedding light on their ability to explain spatial variations in WUE. Initially, a grid system was established over the study area, comprising a total of n grids. In every grid, an overlay analysis was performed to calculate the WUE and the explanatory strength q (ranging from 0 to 1) of the chosen factor, assessing its influence on the spatial variability of WUE throughout China’s drylands. As q approached 1, it indicated a stronger influence of the individual factor on WUE. A q value of 0 signified no relation between the factor and the spatial variations in WUE.

$$q=1-\frac{{\sum }_{h-1}^3 n_{Xh}{\mathrm{V}\mathrm{a}\mathrm{r}}_{Xh}}{n{\mathrm{V}\mathrm{a}\mathrm{r}}_X}$$

(7)

where $n_{Xh}$ denotes the number of subregions influenced by factor X, and ${\mathrm{V}\mathrm{a}\mathrm{r}}_X$ represents the variance of factor X across these subregions.

The Interaction Detector was utilized to compute the q value representing the interplay between the spatial driving factors X₁ and X₂, both of which influence WUE. Following this calculation, the resulting value was compared to the combined sum of the q values associated with X₁ and X₂ when considered independently [26]. The purpose of this comparison was to ascertain if the interaction between X₁ and X₂ amplified or mitigated their joint impact on WUE in China’s drylands, or if their effects were independent (

Table 4. Interactive types between pairs of factors.

Judgment Criteria	Type
q(X1 ∩ X2) < Min(q(X1), q(X2))	Weaken, nonlinear
Min(q(X1), q(X2))) < q(X1∩X2)) < Max(q(X1), q(X2))	Weaken, uni-
q(X1 ∩ X2) > Max(q(X1), q(X2))	Enhance, bi-
q(X1 ∩ X2) = q(X1) + q(X2)	Independent
q(X1 ∩ X2) > q(X1) + q(X2)	Enhance, nonlinear

Notes: q (X₁) and q (X₂) respectively represent the explanatory power of two independent factors.

). Examining how the interaction q value compares to the combined independent q values provided insights into whether X₁ and X₂’s interaction had a synergistic effect on WUE or if their impacts were distinct. This method offered a nuanced understanding of the complex interactions among various factors that shape the spatial patterns of WUE within China’s drylands. The results from the above methods were all at the 99% confidence level.

2.3.6. Hurst Index

The Hurst index (H) can be used to reflect the persistence of the change trend, which has been widely used in the field of ecology. This study used the Hurst index to explore the sustainability of the WUE time series and determine the possible direction of change of WUE in the future. Calculated as follows:

$$H=\frac{\log (R/S)}{\log \left(n\right)}$$

(8)

R is the range of the time series, which is the difference between the maximum value and the minimum value within a period of time. S is the standard deviation, which reflects the degree of fluctuation of the time series during this period. n is the number of observations in the time series. When 0 < H < 0.5, it indicates that the WUE time series has anti-sustainability, and the future change trend is opposite to the actual trend. When 0.5 < H < 1, it indicates that the time series of WUE is persistent, and the future change trend is the same as the actual trend [27].

3. Result

3.1. Temporal and Spatial Variations of WUE

As shown in Figure 2a, the period from 2001 to 2020 demonstrated a general downward trend in WUE across China’s drylands, with a decline rate of −0.004 gCmm⁻¹ H₂O·yr⁻¹. In contrast, both GPP and ET presented increasing trends (Figure 2b), with growth rates of 4.96 gCm⁻²·yr⁻¹ and 4.25 mm·yr⁻¹, respectively. In terms of spatial distribution, WUE displayed a minimal decreasing trend, covering 66.8% of China’s drylands (Figure 2c). These areas were primarily located in the central, eastern, and northeastern drylands of China. Approximately 13.08% of the regions exhibited a significant decreasing trend (p < 0.05) and were primarily distributed in the eastern and northeastern drylands and Xinjiang. The study area exhibited a significant increase in WUE (p < 0.05), accounting for 1.95% of the total scattered sporadically in Qinghai and Tibet (Figure 2c). Approximately 95.37% of the regions displayed a significant increasing trend in GPP (p < 0.05), accounting for 60.71% of China’s drylands. Significant growth was observed in the eastern drylands and parts of Qinghai (Figure 2d). Furthermore, approximately 91.34% of the regions displayed an increasing trend in ET, with 59.3% showing significant growth (p < 0.05). Most of the significant growth in ET was concentrated in the central and eastern study areas (Figure 2e).

Figure 3a illustrates the variability patterns in the WUE, GPP, and ET. The two models with the largest proportion of the study area were C and F (Table 3), accounting for 62.78% and 25.34% of the study area, respectively. More than half of the decrease in WUE was primarily due to the dominant growth of ET (model C), with its growth rate surpassing that of GPP. This pattern was mainly observed in the eastern and central parts of the study area as well as in the southwest edge of Tibet and Xinjiang. Conversely, the increase in WUE was mainly attributed to the growth of GPP (model F), which was concentrated in central Tibet, central Qinghai Province, the border of Inner Mongolia, Hebei, Shaanxi, and Xinjiang. The increase in WUE resulting from the increase in GPP and decrease in ET (model E) was mainly concentrated in northeast Tibet and southwest Qinghai.

The spatial distribution of average WUE in the study area during 2001–2020 was categorized into five groups based on natural discontinuities. As illustrated in Figure 3b, the average WUE in China’s drylands was identified to be 1.15 gCmm⁻¹ H₂O, and the general spatial pattern exhibited an eastward progression from west to east, with the exception of Xinjiang. In Xinjiang, the WUE was higher overall, displaying a spatial trend of decline from the desert edge to the desert center, and was particularly high in the northern region.

3.2. Analysis of Driving Factors on Temporal Dynamics of WUE

The relationships between monthly WUE, GPP, ET, and climatic factors are illustrated in Figure 4. Figure 4a illustrates that, throughout the research period, WUE demonstrated a significantly strong positive relationship with Tem and VPD (p < 0.01), but it did not display any notable correlation with Prec and Solra. Additionally, both gross primary productivity (GPP) and evapotranspiration (ET) were found to have a markedly positive association with precipitation (p < 0.01). ET was significantly negatively correlated with Tem (p < 0.01), whereas GPP was not significantly correlated with Tem. Both GPP and ET were extremely positively correlated with the VPD. In the spring and summer months (Figure 4b,c), WUE displayed a highly significant positive correlation with Prec (p < 0.05). Moreover, in fall (Figure 4d), WUE exhibited an extremely significant positive correlation with Prec and VPD (p < 0.01). Prec demonstrated highly significant positive correlations with GPP and ET during the summer, while exhibiting negative correlations during the spring and fall (p < 0.01). In contrast, VPD showed no significant correlations with GPP and ET in summer but displayed extremely significant positive correlations in spring and fall (p < 0.01) (Figure 4). The effect of Prec on GPP was substantial in all three seasons, with significant positive correlations present in each season, and stronger correlations observed in spring and fall (Figure 4). The partial correlation between GPP and Tem also revealed seasonal differences, with a significantly negative correlation in fall (p < 0.05) and an extremely significant positive correlation in summer (p < 0.01) (Figure 4). The relationship between GPP and VPD was positive in fall but not in other seasons. VPD displayed a positive correlation with ET in fall (p < 0.01) but did not show a significant correlation with ET in other seasons (Figure 4).

The geographic distribution of partial correlations among climatic elements and WUE, GPP, and ET at an annual scale is shown in Figure 5. Over half of the study area displayed a negative correlation between WUE and Prec, among which the eastern drylands of China, central Tibet, and Xinjiang exhibited a strong negative correlation, and the central part of Qinghai Province and the southeastern part of Gansu Province showed a strong negative correlation (Figure 5a). The partial correlation between GPP and ET and Prec was predominantly positive (Figure 5b,c). The strongest positive correlation was observed in the eastern and northeastern regions of the study area (Figure 5b), while the negative correlation between ET and Prec was mainly concentrated in Qinghai and southeastern Tibet. The region with a proportion of 60.53% exhibited a negative correlation between the WUE and Tem (Figure 5d). Additionally, the central part of the study area, as well as the northern parts of Inner Mongolia and Xinjiang, and southwestern Tibet, displayed a strong negative correlation. The partial correlation between GPP, ET, and Tem was primarily positive (Figure 5e,f), with weak positive correlations. The regions with a negative correlation between WUE and VPD were mainly concentrated in the northeastern part of Inner Mongolia and the southern part of Shanxi and Hebei provinces, and sporadic distribution was found in Shaanxi and Gansu regions in the middle of the study area, as well as in Tibet and Xinjiang (Figure 5g). The strongest correlation between GPP, ET, and VPD was observed in the eastern and northeastern parts of the study area and in the northern part of Xinjiang (Figure 5h,i). The partial correlation between WUE and Solra was stronger in the central and Tibetan regions of the study area, while the positive correlation between GPP and ET and Solra was stronger in this region, and the positive correlation between GPP and ET and Solra was stronger in the eastern region, whereas the correlation between WUE and Solra was weaker in this region (Figure 5j–l).

The relative contribution rates of the four climatic factors to WUE, GPP, and ET for the various vegetation types are presented in

Table 5. The contribution rate of climate factors of different vegetation types to the temporal change of WUE, GPP, and ET.

	WUE				GPP				ET
	Prec	Tem	VPD	Solra	Prec	Tem	VPD	Solra	Prec	Tem	VPD	Solra
China’s drylands	25.58%	24.96%	23.54%	25.92%	31.47%	21.82%	22.78%	23.93%	33.31%	24.14%	20.96%	21.59%
Coniferous forest	25.93%	23.63%	24.35%	26.09%	28.18%	23.39%	23.67%	24.76%	29.68%	24.32%	22.34%	23.66%
Mixed forest	25.01%	24.70%	27.63%	22.65%	26.49%	15.95%	26.44%	31.12%	23.02%	22.21%	26.55%	28.22%
Broadleaf forest	26.30%	22.78%	22.80%	28.12%	33.56%	19.88%	23.26%	23.30%	33.04%	21.93%	22.29%	22.74%
Shrub	25.45%	25.16%	24.05%	25.34%	31.57%	20.61%	23.98%	23.83%	32.08%	23.09%	22.11%	22.72%
Desert	25.67%	27.15%	23.28%	23.90%	29.26%	23.37%	26.14%	21.23%	29.53%	26.71%	22.97%	20.79%
Grassland	25.23%	24.73%	23.99%	26.05%	31.35%	22.06%	22.98%	23.61%	34.20%	24.78%	20.57%	20.45%
Swamp	28.34%	24.70%	24.19%	22.77%	23.12%	19.13%	25.33%	32.42%	24.91%	21.60%	25.76%	27.73%
Alpine vegetation	27.42%	24.57%	22.42%	25.59%	23.91%	25.50%	24.80%	25.79%	31.46%	26.21%	21.47%	20.86%
Cultivated vegetation	25.66%	25.47%	22.67%	26.20%	35.14%	20.05%	19.84%	24.97%	33.67%	21.97%	20.25%	24.11%

. The relative contribution rates of Prec to GPP and ET were 31.47% and 33.31%, respectively. The Tem factor had a higher-than-average contribution rate to WUE in desert, shrub, and cultivated vegetation. The contribution rate of forest was lower than the average. Mixed forest had a higher contribution rate than the other forest types. The contribution rate of Prec to GPP was greater in cultivated vegetation, broadleaf forest, shrub, and grassland, whereas it was lower in swamp and alpine vegetation. The contribution rate of Prec to ET in broadleaf forest, grassland, and cultivated vegetation areas exceeded the average, while the contribution rate of Tem was lower than the average. Moreover, the contribution rate of VPD to ET in mixed forest and swamp areas was higher than average, and the contribution rate of Solra to ET in cultivated vegetation, swamp, and forest areas was higher than that in other vegetation types.

3.3. Contribution of Climate Change and Human Activities to WUE Interannual Changes

According to the residual analysis and relative contribution rate method, the contribution rates of climate and human activities to the change in WUE were calculated and are presented in Figure 6. As illustrated in Figure 6a, in Tibet, where the contribution rate of climate change to the increase in WUE was high, as well as in Gansu, Ningxia, and parts of neighboring Inner Mongolia in the central region of the study area, the contribution rate of climate change to the decrease in WUE was significant. WUE primarily decreased in the eastern and northeastern regions of the study area, and the contribution rate of human activities in this area was relatively high (Figure 6b). Additionally, the contribution rate of human activities to the increase in WUE was also notably high in the northern part of Shaanxi Province and southwestern part of Qinghai Province. Human activities played a significant role in the northern region of Xinjiang (Figure 6b). Figure 6 shows the factors responsible for changes in WUE and the percentage of land affected. Over half of the changes in WUE were affected by both climate change and human activities, with 24.78% and 37.79% of the area experiencing increases and decreases, respectively. The area where climate was the sole driver of change (climate contribution rate = 100%) accounted for 8.86% of the total area and was primarily found in Tibet (Figure 6a). On the other hand, the area where human activities were the sole driver of change (human contribution rate = 100%) comprised 7.36% of the total and was primarily concentrated in northern Shaanxi and southwestern Qinghai Province (Figure 6b). Among the regions experiencing an increase in WUE, 8.83% were solely influenced by climate (climate contribution rate = −100%), primarily in northern Xinjiang and Gansu, northeastern Qinghai, and Ningxia (Figure 6a). Meanwhile, 12.57% were solely influenced by human activities (human contribution rate = −100%). The distribution of this particular entity was primarily concentrated in central and northeastern Inner Mongolia, Liaoning Province, and eastern Tibet (Figure 6b).

3.4. Factors Affecting Spatial Differentiation of WUE in China’s Drylands

3.4.1. Dominant Factors of Spatial Differences in WUE

This study selected five periods (2001, 2005, 2010, 2015, and 2020) to determine the driving force behind the spatial differentiation of WUE and calculated the explanatory power (q value) of each factor on China’s drylands’ WUE spatial differentiation. The results of the factor detector are shown in Figure 7. Elev, Geomt, and Soilt all exhibited high explanatory power during the five periods, thereby providing a comprehensive explanation for the primary cause of the spatial differentiation of WUE in China’s drylands. Among the climatic factors, Tem, Solr, Relah, and VPD also showed high explanatory power for the spatial distribution of WUE in China’s drylands. The explanatory power of Relah increased over time, while that of VPD decreased. The explanatory power of Prec as a single factor for spatial drivers of WUE was relatively low (less than 0.2). Simultaneously, GPP had a significant impact on the spatial distribution of WUE. However, the explanatory power of ET was smaller. Among the human activity factors (GDP, POP, and Lanct), Lanct exerted the most significant influence on the spatial variation of WUE, demonstrating a rising trend from 2001 to 2020.

3.4.2. Interactions between Factors Affecting WUE

Figure A1 depicts the size and type of the interactive explanatory power of any two factors, employing bubble size and color. The explanatory power of any two factors across each year surpassed that of a single factor, with the primary types being “Enhance, bi” and “Enhance, nonlinear”. Among these factors, Elev, Geomt, and Soilt demonstrated the strongest interactive explanatory power after interacting with all other factors. Solra possessed the greatest explanatory power among climate factors after interacting with other climate factors, whereas AI exhibited the strongest explanatory power after engaging with Tem and Prec, than to a single factor. Finally, the interactive effect of human activity factors on WUE spatial driving remained relatively modest, with the interaction of land cover and GDP exhibiting the most significant explanatory power.

Table 6. Rank of interactive explanatory power in each year.

Year	Rank of Interactive Explanatory Power (Top Seven)
2001	Soilt∩Geomt = 0.787 > Soilt∩Elev = 0.780 > GPP∩ET = 0.766 > Soilt∩GPP = 0.765 > GPP∩Geomt = 0.760 > Soilt∩Solra = 0.758 > GPP∩Elev = 0.757
2005	GPP∩ET = 0.788 > Soilt∩Geomt = 0.785 > Soilt∩Elev = 0.780 > GPP∩Elev = 0.774 > Soilt∩GPP = 0.771 > GPP∩Geomt = 0.763 > Elev∩NDVI = 0.758
2010	GPP∩ET = 0.786 > Soilt∩Geomt = 0.769 > Soilt∩Solra = 0.758 > Soilt∩GPP = 0.757 > Soilt∩Elev = 0.752 > GPP∩Geomt = 0.745 > GPP∩Elev = 0.741
2015	GPP∩ET = 0.797 > Soilt∩Geomt = 0.785 > Soilt∩Elev = 0.778 > Soilt∩GPP = 0.775 > GPP∩Elev = 0.761 > Soilt∩Solra = 0.760 > GPP∩Geomt = 0.753
2020	GPP∩ET = 0.789 > Soilt∩Geomt = 0.784 > Soilt∩GPP = 0.779 > Soilt∩Elev = 0.775 > GPP∩Elev = 0.761 > GPP∩Geomt = 0.755 > Relah∩Soilt = 0.744

depicts the seven combinations with the most potent explanatory power for the two factors across the five time periods. Except for the strongest explanatory power of GPP and ET, the interactions between Geomt, Soilt, GPP, Solra, and NDVI exhibited the strongest explanatory power for the spatial differentiation of WUE in China’s drylands.

3.5. Future Trends in WUE Dynamics

The Hurst index (H), as depicted in Figure 8a, elucidated the endurance of WUE alterations. In this regard, H was classified into four distinct categories. In China’s drylands, the WUE of an extensive area encompassing 78.12% of the region was characterized by anti-persistence. Notably, strong anti-persistence accounted for 16.91% and was primarily distributed in Tibet, with occasional occurrences in the eastern regions. To determine future changes in WUE, the Hurst index was compared with the current trend of WUE in a specific spatial region, resulting in eight possible scenarios, as illustrated in Figure 8b. The future change in WUE was primarily characterized by an increase, accounting for 56.98% of the total. These regions were primarily located in the eastern, northeastern, and northern parts of Xinjiang of the study area. Additionally, 43.02% of the regions experienced a decrease in WUE, with these areas primarily situated in Qinghai and Tibet. Finally, it was reported that 6.98% of the regions showed a strong and continuous decrease in WUE, which was particularly prominent in Tibet.

4. Discussion

4.1. Temporal and Spatial Variations in WUE and Attribution

Excluding the Tibetan Plateau region, where WUE was rising, the other regions exhibited varying degrees of decline (Figure 2). GPP displayed an ascending trend in the majority of the study area, suggesting that the accumulation of ecosystem organic matter in China’s dryland ecosystems was on a rising trend during the study period, and the impact of vegetation restoration was evident. As demonstrated in the superposition analysis (Figure 3), the reduction in WUE across 62.78% of the total area was primarily attributable to an increase in ET, whereas the decrease in WUE in the Tibetan Plateau region was mainly attributed to an increase in GPP. Due to the fluctuating nature of climate change, there is also some uncertainty in the direction of WUE change in different regions of the dry zone over different time periods. However, whatever the changes are, they are the result of certain climatic fluctuations and human activities.

Table 7. Trends in WUE in related studies.

Area	Time	WUE Trend
Inner Mongolia [44]	2001–2020	Decrease
Tibetan Plateau [45]	1979–2010	Increase
Three-North Region of China [46]	2001–2017	Increase

shows the trends of WUE in some of the previous research results.

Furthermore, as indicated in a study by Zhang et al., both WUE and GPP can be utilized to evaluate the process of vegetation restoration [32]. During the early phase of vegetation restoration, biomass accumulation, vegetation cover enhancement, and concurrent increases in the WUE were observed. As vegetation restoration progressed to its later stages, plant stomatal conductance increased, drought-tolerant plant coverage declined, and WUE may have decreased [39,47]. In the majority of China’s arid regions, excluding a significant portion of the Qinghai–Tibet area, the increase in GPP was accompanied by a reduction in WUE during the observation period, indicating that vegetation recovery in these regions was at a more advanced stage. On the other hand, the occurrence of enhanced WUE along with increased GPP in certain areas of the Tibetan Plateau indicated that vegetation recovery in this region may be in its initial stage.

4.2. Sensitivity of WUE and GPP to Climate Change

The reactions of WUE and GPP to climatic variables varied across both monthly and yearly time frames. At the monthly scale, Prec and VPD were the primary drivers of the monthly changes in GPP (Figure 4). Prec served as the primary source of surface soil water in the drylands, which in turn was the main driver of drought stress in the ecosystems [48], suggesting that drought stress had a strong impact on plant organic matter production in the drylands. In terms of the degree of response, the correlation coefficients for spring and fall were slightly larger than those for summer, indicating that Prec in spring and fall had a more significant effect on increasing GPP, which was primarily due to the fact that Prec in the drylands was mainly concentrated in the summer and mostly occurred in the form of storm rainfall. The effect of water stress on vegetation growth was more pronounced during the spring and fall seasons when Prec was sparse [49]. Fall Prec was observed to improve WUE, whereas spring Prec was associated with a decline in WUE. Previous studies have also indicated that the enhancement of spring rainfall was a key factor in promoting vegetation regeneration in arid regions, which played a crucial role in revitalizing the western drylands [29,50]. Climatic factors exhibited distinct responses to WUE across different seasons (Figure 4), suggesting that vegetation employed varying water utilization strategies at different developmental stages. Throughout the year, the two meteorological factors that had the most significant impact on fluctuations in WUE were Tem and VPD, and both exhibited a positive correlation. Tem showed a highly significant negative correlation with ET (p < 0.01), while it did not demonstrate a significant correlation with GPP. This suggested that the influence of Tem on monthly changes in WUE was primarily realized through its effect on ET.

On the scale of interannual variability, except for the cold drylands of the Tibetan Plateau, Prec had the highest contribution to the increase in GPP among the four factors, amounting to 31.47%, and showed a strong positive correlation. The relative importance of the other three climatic factors to GPP was ranked as Solra>VPD>Tem, indicating that the increase of Prec was an important reason for the increase of GPP in the drylands of China. Moreover, the increase of Prec enhanced the effectiveness of water and thus GPP, and also verified the conclusion that the role of water stress in the drylands was significant [49]. At the same time, the Prec was one of the most important factors affecting the ET in the study area, and the contribution rate was 33.31%, which this indicates that the increase of ET was closely related to the increase of rainfall. Meanwhile, the negative correlation between Prec and WUE was dominant, indicating that in the drylands of China, the increase of Prec promoted the growth of GPP and increases the ET at the same time to lead to the decrease of WUE. Moisture changes had a more significant effect on GPP than heat, and the potential for vegetation to fix carbon was greater when the Tem reached the requirements for vegetation growth and development, so GPP increased significantly when Prec increased. This aligned with findings from earlier research [51,52,53]. Solra predominantly showed a positive correlation with both WUE and GPP, indicating that Solra mainly acted on WUE by affecting the fixation of carbon in plants (Figure 5). Zheng et al. pointed out that [54], under drought conditions, the WUE negative sensitivity to VPD was obvious, and our study found that VPD was predominantly negatively correlated with both WUE and GPP in the drylands of China, suggesting that an increase in VPD can reduce carbon assimilation and water use in the ecosystems under water stress [55]. Although the scientific validity of these conclusions has been confirmed, we inevitably acknowledge certain limitations. In exploring the impact of climate change on WUE, our analyses have not yet taken into account some of the independent variables that may be present in the time series of the relevant factors. This becomes a matter of concern for us in the future.

4.3. Role of Human Activities in the Temporal Variation of WUE

This study utilized residual analysis and relative contribution rate calculation methods to assess the contributions of climate and human activity to changes in WUE in China’s drylands. The findings indicated that human activities contributed to 46.35% of WUE changes, whereas climate change accounted for 53.65%. Human activities had a relatively dominant influence. In terms of spatial distribution, more than half (62.57%) of WUE changes were attributed to both climate change and human activities. The study area showed a concentration of larger anthropogenic contributions in the eastern and northeastern parts of China’s drylands, where human activities were stronger, which aligned with the findings of Li Ming et al.’s study on WUE in Inner Mongolia [21]. Since 1978, China has implemented 13 protection and restoration projects, such as the “Grain to Green Program” and the “Three-North Shelterbelt Program” [29], which were proven to be effective in greening China’s drylands and partially responsible for the increase in GPP in the study area (Figure 2d). Furthermore, human activities in large areas of the eastern and northeastern parts of the study area, including the Xilingol League in Inner Mongolia, Hebei Province, and eastern Liaoning Province, have contributed to a decrease in WUE (Figure 5b). On the one hand, the expansion of urban development and land use [21] in these areas can lead to a reduction in green space, and the ground hardening caused by urbanization can increase ET and reduce WUE. On the other hand, large-scale afforestation was vital for increasing ET. It has been observed that ecological engineering measures, such as tree planting, can result in lowering of the water table and drier deeper soils [56,57], and afforestation activities can cause significant changes in local plant communities. Although these plant communities can contribute to the greening of drylands, they can also result in a significant increase in evapotranspiration from the ecosystem.

4.4. Factors in the Spatial Drivers of WUE

The trend of the spatial distribution pattern for WUE was consistent with the findings of previous studies [20]. The results of the factor detection analysis revealed that the top three factors contributing to the spatial variation in WUE in the five time periods were Elev, Geomt, and Soilt. These findings indicated that topographic factors and soil conditions played a major role in the spatial variation in WUE in the drylands of China. Different soil types possessed varying soil water contents, soil organic matter, and soil textures. Previous studies have demonstrated that soil water content and organic matter are crucial drivers of WUE [50]. The application of organic matter was a significant factor in determining WUE, and the combined influence of these factors rendered WUE more sensitive to Soilt. Meanwhile, the results of the interaction detection demonstrated that the factors Geomt and Soilt, as well as Soilt and Elev, had the greatest explanatory power, which suggested that the integrated conditions of topography and soil were the primary drivers of the spatial differentiation of the WUE driving force. In addition, among the climatic factors, Tem, Solra, Relah, and VPD had more substantial impacts on WUE. However, Prec did not emerge as the primary determinant of spatial WUE divergence in China’s drylands, indicating that the influence of the Tem gradient on GPP and ET may offset the moisture stress-induced spatial WUE [20]. Among the anthropogenic factors (GDP, POP, and Lanct), Lanct appeared to exert the most significant impact on WUE, and there was a discernible upward trend over time, suggesting that the impact of human activities on the spatial disparities in WUE was progressively intensifying. However, we recognize that the interactions among the dominant factors of spatial differentiation are subject to further future research.

4.5. Future Trend in WUE

Most of the regions (78.12%) in the study area exhibited inverse sustainability in terms of changes in WUE. In eastern areas, there would be a possibility of enhancing WUE. However, alpine meadow grassland areas in Tibet may experience a decrease in WUE in the future. WUE increased within a specific range to alleviate water stress in drylands. GPP in the Tibetan Plateau region continued to rise in the context of vegetation greening, but further vegetation greening could lead to increased evapotranspiration and exacerbate soil aridity. Research has indicated that ecosystems may face challenges in maintaining their current state when faced with persistent soil moisture depletion, which could result in reduced productivity, vegetation degradation, and a decrease in WUE [29,58]. Therefore, it is recommended that the vegetation dynamics of the Tibetan Plateau should be prioritized in future studies, and restoration strategies should be developed to prevent further degradation of the vegetation.

5. Conclusions

This study represents the first comprehensive temporal and spatial analysis of the impacts of climate change and human activities on WUE in China’s drylands, revealing how these factors collectively influence GPP and ET. The interactions among WUE, GPP, and ET are illustrated in Figure 9. Through an exhaustive analysis of data from 2001 to 2020, we found a significant increasing trend in GPP across China’s drylands, while WUE demonstrated fluctuating declines. The interannual variability in WUE was primarily attributed to the dual influences of climate change and human activities, with Solra and Prec identified as key climatic drivers. Human activities, especially in the eastern regions, significantly contributed to the reduction in WUE, while geomorphological and soil conditions were the main factors affecting spatial variability in WUE. Our research underscores the need to pay attention to the potential risk of vegetation degradation against the backdrop of global warming and increasing humidity in drylands, particularly in the Tibetan Plateau. Continuous monitoring and research into the dynamics of WUE in China’s drylands are thus crucial. Our findings highlight the delicate balance between ecosystem productivity and water usage, emphasizing the need for ongoing research and monitoring efforts to effectively address environmental challenges in these regions.

Future studies should focus on the interaction between WUE and GPP and its long-term impact on ecosystem health. This is vital for optimizing future ecological engineering projects in China’s drylands. This study not only provides new insights into understanding and managing the water resources and ecosystems of China’s drylands but also offers a reference for the ecological restoration and sustainable management of similar arid regions globally.