Drought is a common hydroclimatic hazard and has unfavorable consequences for terrestrial ecosystems (Mishra and Singh, 2010; Vicente-Serrano et al., 2020). Drought has been recognized as a critical component in the circulation between terrestrial water and carbon (Gentine et al., 2019), affecting soil evaporation, plant transpiration, and carbon absorption at the ecological scale (Liu et al., 2019; Yang et al., 2021). Climate model projections suggest that as a consequence of climate change, the frequency and intensity of droughts will increase (Dai, 2012; Su et al., 2021) and are expected to pose extensive and adverse impacts on ecosystem carbon and water exchange (Huang et al., 2016; Gentine et al., 2019; Zhang et al., 2020). The water use efficiency (WUE) is calculated as the rate of carbon assimilation during photosynthesis to water loss via transpiration, and it is a crucial metric of the coupling of carbon and water in the ecosystem. The ratio of the evapotranspiration (ET) to the gross primary production (GPP) at the ecosystem scale can be used to determine the WUE (Huang et al., 2016; Kang and Kang, 2019). Knowledge of drought impacts on WUE is essential for elucidating how terrestrial ecosystems respond to climate change (Yi and Jackson, 2021; Zhao et al., 2021; Zhou et al., 2021).

The WUE has been extensively inspected on the basis of site measurements (Acuna et al., 2015; Medrano et al., 2015; Lin et al., 2018; Duan et al., 2020). Scientists have examined the spatial variability of the WUE in various regions using network eddy covariance measurements (Zhu et al., 2015; Zhao et al., 2021). However, due to the restricted observations of the eddy covariance flux towers, an accurate estimation of the ecosystem WUE at the global scale remains difficult (Yang et al., 2020). Recent research has shown that process-based ecosystem models and remote sensing methods provide new tools to help us comprehensively understand the WUE dynamics on regional and even global scales (Wei et al., 2019; Yang et al., 2020, 2021). For example, the moderate resolution imaging spectroradiometer (MODIS), breathing Earth system simulator (BESS), Penman-Monteith-Leuning model (PML), Global Land Surface Satellite (GLASS), and global land evaporation Amsterdam model (GLEAM) have been extensively utilized to assess the WUEs of regional or global terrestrial ecosystems (Liu et al., 2019; Xu et al., 2019; Zhao A. et al., 2020; Measho et al., 2021; Yang et al., 2021).

WUE variations are strongly affected by various biotic (i.e., leaf area index and vegetation types) and abiotic factors (i.e., precipitation, temperature, and radiation) (Liu et al., 2015). The influencing factors of WUE had significant spatial heterogeneity and varied with the time scales. Li et al. (2023) employed structural equation modeling (SEM) to investigate the influences of WUE in a subtropical evergreen forest in southeastern China and found that WUE responded significantly to soil moisture (SM) at different time scales (from half-hourly to monthly scales). Qin et al. (2023) used geographic detector modeling (GDM) to discover that land surface temperature (LST) had the strongest driving effect on WUE in Central Asia, and temperature (TEM) had the largest interaction effect with vegetation cover. Nandy et al. (2022) used a random forest (RF) algorithm to reveal that TEM is the main driver of the WUE in Indian forests.

Drought may simultaneously alter both components of the WUE (the GPP and ET) and is therefore regarded as one of the most influential abiotic variables affecting the WUE (Zhao A. et al., 2020). The impacts of drought on ecosystem WUE vary according to the vegetation types, drought severity, and dry and wet zones (Zhao J. et al., 2020). According to Yang et al. (2021), over 50% of the vegetated regions worldwide have positive correlations between the WUE and drought, whereas the negative correlations basically were distributed in the high-latitude areas. Yang et al. (2016) discovered that there are strong negative relationships between the WUE and the drought index in dry areas, while sub-humid and semi-arid zones have positive relationships. This discovery is in line with the results of Huang et al. (2017), who demonstrated that the WUE in wet ecosystems often exhibited both a positive and negative response to drought, while the WUE in dry ecosystems frequently exhibited a negative response to drought. However, Guo et al. (2019) reported a positive correlation between the WUE and the standardized precipitation and evapotranspiration index (SPEI) in subarid China. In areas with a high humidity, the response of the WUE drought indices was found to be both positive and negative.

The core concept for determining how an ecosystem responds to a major disturbance is based on resilience (Yi and Jackson, 2021). The ecosystem resilience mainly describes the state changes in the ecosystem before and after the response to interference (Ingrisch and Bahn, 2018). Malone et al. (2016) quantified the resilience of the ecosystem to drought by comparing the changes in the WUE in normal years and under drought conditions. Their results revealed that the arid ecosystem in California has more drought resistance than the high-productivity ecosystem in the northern part of the state. Sharma and Goyal (2018a) introduced a resilience index and found that many of India’s river basins were insufficiently resilient to drought disturbances. Guo et al. (2019) reported that certain ecosystems in northeastern, central, and southwestern China demonstrated relatively high levels of resilience based on the dimensionless resilience index developed by Sharma and Goyal (2018a). In addition, Liu et al. (2019) further assessed the WUE resilience to drought in China in several dimensions, including the grid cell, land use, climate type, river basin, and province.

The Lancang–Mekong River Basin (LMRB) is the largest cross-border river basin in Southeast Asia, and is unparalleled in terms of its biodiversity and ecosystem services. Drought events have been more common in the LMRB during the last few decades. Examples of years with extreme droughts that threatened water and grain security include 1992–1993, 1998–1999, 2003–2005, 2010–2011, 2015–2016, and 2019–2020 (Guo et al., 2017; Tran et al., 2019). Recent research indicates that both meteorological and hydrological drought will intensify in the near future (Li Y. et al., 2021). Nonetheless, there is a lack of comprehensive analysis regarding how the WUE responds to drought occurrences, as well as a deficiency in research pertaining to ecosystem resilience to drought, which is the primary natural disaster in the LMRB. Therefore, understanding the impacts of drought on the WUE is imperative for managing drought risks and minimizing uncertainties in forecasting future terrestrial carbon sequestration. We evaluated the response of the WUE to drought in the LMRB from 1982 to 2018 utilizing remote sensing and model-based data. The particular aims of this study were (1) to investigate the spatiotemporal changes in the GPP, ET, and WUE from 1982 to 2018; (2) to explore how the WUE responded to drought and the ecosystem resilience; and (3) to examine the changes in WUE and ecosystem resilience across different land use types, climate types, and elevation gradients.

On the basis of the GLASS GPP and ET, we estimated that the mean yearly WUE across the LMRB during 1982–2018 was 1.63 g C kg⁻¹ H₂O. Using a similar method, Wang L. et al. (2020) and Pokhrel et al. (2021) analyzed the ecosystem WUE of the surrounding regions and found that the multi-year averaged WUEs of Tibetan Plateau and China were 0.65 g C kg⁻¹ H₂O and 1.08 g C kg⁻¹ H₂O, respectively. They additionally reported an increase in the WUE over the study period in this region. Contrary to their findings, we found that the yearly WUE in the LMRB declined at a rate of −0.0009 g C kg⁻¹ H₂O·yr⁻¹. The decline in the WUE in the LMRB was caused by an increase in the GPP, which was largely offset by an increase in ET under global warming.

The findings of our study indicate that TEM and VPD were important factors affecting the spatial distribution pattern of WUE in LMRB (Figure 5). Nandy et al. (2022) found similar results in India, where TEM was the most important driver of the WUE in forests. Jiang et al. (2019) also discovered that these two factors in the Pacific Northwest mainly controlled WUE on some short time scale. The mean yearly WUEs of the various types of vegetative cover differed, with forests having a greater WUE than grasslands and croplands having the lowest WUE (Figure 9). These findings are consistent with the global and Chinese trends of the average yearly WUE reported in prior scholarly works (Tang et al., 2014; Li G. et al., 2021). Nevertheless, other investigations conducted in China have reported that grasslands exhibited the lowest WUE values (Liu et al., 2015; Guo et al., 2019). The variation in the outcomes of these studies may be partially ascribed to the differences in the environmental circumstances, plant varieties, cultivation regions, and agronomic practices. In addition, the WUEs of the warm temperate climate types (Cfb, Cwa, Cwb, and Csb) were higher, while those of the cold and polar climate types (Dwe, Dwd, and EFH) were lower (Figure 10), which is consistent with previous finding at the global scale (Xia et al., 2015) and in India (Sharma and Goyal, 2018a). In arid ecosystems, the evaporation of water from bare soil, which is not necessarily linked to productivity, has a tendency to reduce the WUE in comparison to more humid regions (Sun et al., 2016). Overall, the WUE exhibited a decreasing trend that initially increased and then decreased with increasing elevation. The highest value occurred in the 2,500–3,000 m elevation gradient (2.69 g C kg⁻¹ H₂O) (Figure 11). These results are consistent with the results of Xue et al. (2015), that is, an inverse connection between the WUE and altitude across a global range of 1,150–5,000 m. Negative correlations were also discovered between the elevation gradient and the WUE in China (Zhu et al., 2015) and India (Nandy et al., 2022). The variations in the vegetation composition and climatic conditions with elevation gradients gave rise to distinct differences in the WUE at low and high elevations (Liu et al., 2020).

The spatial heterogeneity of the correlation between the WUE and drought was noteworthy and was attributable to the diverse vegetation, climate, and underlying surface properties. The positive correlations were mostly observed in the humid climate zones, while the negative correlations primarily occurred in the cold and dry regions (Figure 7). This result is in line with the findings of previous studies (Yang et al., 2016, 2021; Huang et al., 2017). In high-latitude regions and semi-arid/sub-humid zones, the correlation between the WUE and drought was influenced by the ET, while in tropical forests, it was mainly determined by the GPP (Yang et al., 2021). Plants in arid areas respond to moisture limitations more rapidly, with larger decreases in the ET than in the GPP, increasing the WUE (Guo et al., 2019). In contrast, plants in humid environments are less able to adjust to drought, so the dry conditions had a larger influence on the plant WUE (Zhao et al., 2021). Additionally, the lagged effect of drought on the WUE has been reported in previous studies (Liu et al., 2015; Yang et al., 2016; Huang et al., 2017; Ji et al., 2021). Our results indicate that the ecosystem WUE lagged behind the drought by 3 months in most regions of the LMRB. The lagged effect was more apparent in the grassland-dominated upstream region and cropland-dominated Mekong Delta. Comparable results were obtained by Liu et al. (2015), who revealed that the lagged influence of drought on the monthly WUE was relatively longer in regions in China that were dominated by farmland and grassland and shorter in regions that were predominantly forested.

Our results show that 38.90% of the total area was resilient to droughts (Figure 8). However, the ecosystem resilience varied with the land use, climate types, and elevation gradients. Regarding land use type, the ecosystem resilience of the CNV exhibited strong drought resistance, while the GR was less resistant to droughts (Figure 9), which is in line with the findings of Tian et al. (2011) and Guo et al. (2019), who reported that grassland is more vulnerable to drought in China and monsoonal Asia, and the main for this reason is that grassland has a low water storage capacity and shallow roots, so it is easily affected by fluctuations in the shallow soil water (Liu et al., 2021; Zhao et al., 2021). Regarding the climate type, the temperate climate types (Csb, Cwb, and Cfb) had a higher resilience, while the ecosystems in the cold and polar climate (Dwc) were severely non-resilient, which is consistent with the findings of Sharma and Goyal (2018a,b). Regarding elevation gradient, the areas above 4,500 m were less resistant to drought, and the vegetation composition and climate conditions in the high elevation area gave rise to poor resilience to drought.

In addition to climate change, previous studies have concluded that human activities also contribute to WUE changes. For example, Zheng et al. (2019) found that the main driver of the rising annual WUE of the ecosystem on the Chinese Loess Plateau is due to revegetation. According to Du et al. (2019), human activities, including afforestation, were responsible for around 16% of the WUE trends in northern China. On the basis of the findings of Wang L. et al. (2020), while human activities have had a negative impact on the WUE, with a reduction of 20.2%, the increase in the WUE on the Tibetan Plateau was primarily driven by climate change, with a contribution rate of 78.9%. Li G. et al. (2021) reported that the rise in the vegetation WUE in China from 2001 to 2014 was driven by human activities (66.75%) and climate change (33.25%). As a result, further investigation is warranted to examine how human activities and climate change will impact the WUE in the LMRB in the future.

It should be noted that our study still had several uncertainties. Even though the GLASS GPP and ET performed rather well in terms of spatial pattern and temporal dynamic (Zhao et al., 2013; Liang et al., 2021), the WUE trends calculated from GLASS had some discrepancies with other remote sensing-driven models (Yang et al., 2021). Additionally, high-resolution precipitation and PET information would surely aid in the monitoring of droughts. However, both satellite-based and model-based products were subject to inherent errors. For instance, the CHIRPS precipitation had high correlation coefficients with gauge observation, however, it tended to systematically overestimate precipitation in the upstream region and underestimate precipitation in the downstream region of the LMRB (Luo et al., 2019; Wu et al., 2019). The GLEAM-PET also was associated with errors due to the calculation methods and forcing data (Jiang et al., 2021). Moreover, the spatial interpolation of datasets with various spatial resolutions (e.g., the original spatial resolutions of GLASS, ERA5-land, and GLEAM are 0.05°, 0.1°, and 0.25°, respectively) may also cause uncertainties. To lessen uncertainty brought on by a single model or dataset, additional efforts must concentrate on integrating multiple datasets to better understand ecosystem WUE response to drought.

In this study, the spatiotemporal variations in the WUE and its response to drought in the LMRB from 1982 to 2018 were investigated. The changes in the WUE and ecosystem resilience to drought were then assessed in relation to land use types, climate types, and elevation gradient. The main findings of this study are as follows:

(1) The mean yearly WUE was 1.63 g C kg⁻¹ H₂O in the LMRB, and the forests and warm temperate climates had higher WUE values. The interaction of temperature and leaf area index were the main factor affecting the spatial distribution of WUE. The yearly WUE exhibited a decreasing trend with a rate of −0.0009 g C kg⁻¹ H₂O·yr⁻¹ across the entire region, and 63.05% of the total area experienced a spatially decreasing WUE trend.

(2) The annual WUE was positively correlated with drought in the humid region, accounting for 66.55% of the total area, while it mainly exhibited a negative relationship in the high-altitude cold region.

(3) The ecosystem WUE lagged behind the drought by 3 months in most regions of the LMRB. The lagged effect was more apparent in the grassland-dominated upstream region and cropland-dominated Mekong Delta.

(4) The results of the resilience analysis suggest that the ecosystems in forests and temperate climates were strongly resistant to drought, while the grassland and high-altitude regions with dry and cold climates had a poor resilience.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

XL: investigation, methodology, formal analysis, and writing–review and editing. YW: data curation, software, and investigation and review. YL: supervision, conceptualization, writing–review and editing, project administration, and funding acquisition. All authors contributed to the article and approved the submitted version.

This work was sponsored by the National Natural Science Foundation of China (Grant No. 42067033) and Applied Basic Research Programs of Yunnan Province (Grant Nos. 202001BB050073 and 202101AT070075).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.