Climate Change Contributions to Water Conservation Capacity in the Upper Mekong River Basin

Abstract

Investigations into the impacts of climate change on water conservation capacity in the upper Mekong River Basin (UMRB) are important for the region’s sustainability. However, quantitative studies on isolating the individual contribution of climate change to water conservation capacity are lacking. In this study, various data-driven SWAT models were developed to quantitatively analyze the unique impact of climate change on water conservation capacity in the UMRB. The results reveal the following: (1) From 1981 to 2020, the annual water conservation capacity ranged from 191.6 to 392.9 mm, showing significant seasonal differences with the values in the rainy season (218.6–420.3 mm) significantly higher than that in the dry season (−57.0–53.2 mm). (2) The contribution of climate change to water conservation capacity is generally negative, with the highest contribution (−65.2%) in the dry season, followed by the annual (−8.7%) and the rainy season (−8.1%). (3) Precipitation, followed by evaporation and surface runoff, emerged as the critical factor affecting water conservation capacity changes in the UMRB. This study can provide insights for water resources management and climate change adaptations in the UMRB and other similar regions in the world.

1. Introduction

Watersheds are essential for water resource management, playing a key role in storing, allocating, and delivering water. This is crucial for ensuring water supply, preserving ecosystem health, and supporting socio-economic development. Climate change attributed to human activities significantly impacts the water cycle [1], exacerbates water scarcity [2,3,4] and extreme precipitation [5], and threatens ecosystem services [6,7,8] on a watershed scale. Notably, the water conservation capacity is a critical ecosystem service function in watersheds and is instrumental in maintaining the sustainability of water resources and ecological equilibrium [9].

Investigating water conservation capacity on a watershed scale can enhance a comprehensive understanding of natural processes, such as precipitation, snowmelt, evaporation, runoff, and groundwater recharge, thereby facilitating a more nuanced grasp of water conservation dynamics. Jia et al. [10] examined the effects of the ecological retreat project on water conservation in the Yellow River Basin. Wu et al. [11] delineated the annual, monthly, and daily water conservation capacity in the Heihe River Basin. Zhang et al. [12] estimated the water conservation capacity in the Weihe River Basin and measured the capacity after removing the impacts of precipitation and evapotranspiration.

Many methodologies, mainly including hydrological experimentation, remote sensing, and hydrological modeling, can be used in water conservation capacity assessment. Hydrological experiments require extensive time and manpower, limiting their large-scale applicability [13,14]. Remote sensing allows for macro-scale quantitative analysis with expansive coverage, short operational periods, and cost-effectiveness, but it cannot fully explain underlying mechanisms and processes [15,16]. Consequently, hydrological models, such as InVEST, WEPL, and SWAT [17,18,19], have become pivotal tools in estimating water conservation capacity. And SWAT models are gaining widespread recognition for their robust performance globally. For instance, Zhao et al. [20] developed a SWAT model to analyze water conservation variations across diverse ecosystems within the source region of the Yellow River. Abouabdillah et al. [17] employed a SWAT model to evaluate the effects of dam construction and soil and water conservation strategies on the water balance and soil erosion within Tunisia’s Merguellil Basin. Wang and Cao [21] applied a SWAT model to quantify the water conservation in the upstream Xiong’an region and further dissect the determinants influencing the spatiotemporal dynamics of water conservation capacity.

The Mekong River, originating from the Qinghai–Tibet Plateau in China, flows through Myanmar, Laos, Thailand, Cambodia, and Vietnam. It is a crucial international river, supporting millions of livelihoods and ensuring water resource security and environmental sustainability in Southeast Asia and globally [22]. The upper Mekong River Basin (UMRB), above Thailand’s Chiang Saen station, constitutes the principal source of the river’s water resource [23]. Water conservation capacity is intricately linked to precipitation, evaporation, and runoff, and it is influenced by a sequence of key factors. Climate change is a prominent factor as it directly alters the water resource volume and the hydrological cycle within a basin [24]. Land use and vegetation cover also play significant roles in shaping surface runoff [25]. Additionally, topographic characteristics, including elevation and slope, contribute to the water conservation dynamics [26]. Human activities, such as dam construction and agricultural practices, further influence these processes [27]. In the UMRB, where alterations in land use and topography are not significant, climate change (increasing temperatures and precipitation uncertainty) and human activities (dam construction) stand out as the predominant influences, impacting downstream agriculture and ecosystems [28,29]. Investigating the effects of climate change on water conservation in the UMRB is crucial for maintaining ecological balance and enhancing water resource security, agricultural productivity, and environmental safety in Southeast Asia.

Previous studies have mainly focused on how climate change (e.g., variations in precipitation and evapotranspiration) and human activities (e.g., dam construction and land use changes) impact the hydrological processes across the UMRB. These studies, using methods like distributed hydrological models, the Budyko hypothesis, and remote sensing, have concentrated on runoff and sediment load to quantify the contributions of climate change and human activities to changes in watershed water resources [22,29,30,31,32,33,34,35]. A few studies extend beyond runoff and sediment to explore other factors; for instance, Bibi et al. [36] utilized GRACE and GLDAS to examine the impact of climate change on terrestrial water storage, while Gui et al. [37] developed an assessment framework to dynamically analyze the vulnerability of the water resource system. A comprehensive analysis of the mentioned studies reveals that none of these studies have calculated the water conservation capacity of the UMRB nor have they thoroughly investigated the mechanisms by which climate change affects water conservation. What are the spatiotemporal evolution patterns of water conservation in the UMRB over the past forty years? How do climate change and its unique contributions affect it? These remain unclear.

Therefore, this study constructed SWAT models to (1) delineate the spatiotemporal evolution of water conservation capacity in the UMRB from 1981 to 2020; (2) isolate and characterize the unique contribution of climate change to the water conservation capacity in the UMRB; and (3) unravel the underlying mechanisms and processes by which climate change influences the water conservation capacity. Additionally, the Mann–Kendall test method was used to detect trends and change points in water conservation. These contributions clarify how climate change specifically affects water conservation in the UMRB and explore how advanced management strategies and technological innovations can enhance the region’s water conservation capacity. This study exemplifies global water resource management by presenting a comprehensive methodology that not only isolates the effects of climate change on water conservation but also assesses the key factors that influence it. Its scalability and adaptability make it applicable to a wide array of hydrological settings, facilitating a robust and proactive approach to climate change adaptation and sustainable water resources use worldwide. This study proposed innovative perspectives and methodologies which can enhance global water resources management, advocating for sustainable water utilization and conservation.

2. Study Area

Originating from the Qinghai–Tibet Plateau in China, the Mekong River is an important transboundary river in Asia with an average annual flow of 14,500 m³/s [38], a length of 4909 km and an area of 795,000 km² [39]. The UMRB, extending north of Thailand’s Chiang Sean station and located between 20° N–34° N and 93° E–102° E, is predominantly situated within China and comprises about 23% of the basin’s total area (Figure 1). In the high-altitude regions of the Qinghai–Tibet Plateau, the climate is classified as a plateau climate. As the basin extends into Yunnan Province, it transitions to a subtropical monsoon climate [40], with average temperatures ranging from 0 to 20 °C and an annual precipitation between 500 and 1600 mm.

The rainy season is from May to October which accounts for 85% of annual precipitation, and the dry season is from November to April [41]. The main vegetation types include alpine, meadow, shrub, and various forests, and the predominant soil types are gelic leptosols, ferric acrisols, and orthic acrisols [33].

3. Materials and Methods

3.1. Data Collection and Pre-Processing

The daily observational meteorological data from 27 stations within the UMRB were obtained from the China Meteorological Administration. The dataset encompasses daily precipitation, daily maximum and minimum temperatures, sunshine hours, relative humidity, and wind speed. Land use data were sourced from Zhang et al.’s [42] global land-cover product (GLC_FCS30-2020) with 30 m spatial resolution. The study incorporated land use data for the years 1985, 1995, 2005, and 2015, segmented into seven categories: agricultural land, forest, pasture, wetland, urban areas, barren land, and water. Soil data were derived from the HWSD [43], which is provided by the United Nations Food and Agriculture Organization. DEM data with 90 m spatial resolution were obtained from the Geospatial Data Cloud site, Computer Network Information Center, Chinese Academy of Sciences. Hydrological data for the Chiang Sean station were sourced from the Mekong River Commission. The basin and administrative boundary data were based on the China Surveying and Mapping Bureau’s information.

3.2. Water Conservation Calculation Method

The hydrological response unit (HRU) is the fundamental unit within the SWAT model. The water balance in an HRU can be given as follows [20]:

$${SW}_{ij}={PRECIP}_{ij}-{ET}_{ij}-{WYLD}_{ij}$$

(1)

$${WYLD}_{ij}={SURQ}_{ij}+{LATQ}_{ij}+{GWQ}_{ij}-{TLOSS}_{ij}$$

(2)

where i is the ith HRU; j is the temporal dimension; SW_ij is the variation in water volume for the ith HRU at the time interval j, mm; PRECIP_ij is the precipitation, mm; ET_ij is the actual evapotranspiration, mm; WYLD_ij is the water yield, mm; SURQ_ij is the surface runoff, mm; LATQ_ij is the lateral flow, mm; GWQ_ij is the groundwater, mm; and TLOSS_ij is the transmission loss, mm.

Based on SWAT, the water conservation capacity at the HRU level can be calculated as follows [20]:

$$\begin{gathered} {WC}_{ij}={SW}_{ij}-{TLOSS}_{ij}+{LATQ}_{ij}+{GWQ}_{ij} \\ ={PRECIP}_{ij}-{ET}_{ij}-{SURQ}_{ij} \end{gathered}$$

(3)

where WC_ij is the water conservation capacity of the ith HRU at the time interval j, mm.

3.3. Unique Contribution of Climate Change

The control variable method is utilized to delineate the unique contribution of climate change to water conservation capacity in this study. This method entails maintaining land use data as constant, while solely modifying meteorological data, thereby ensuring that any resulting variations in water conservation capacity are exclusively attributable to climatic variations. This study formulated eight distinct scenarios, detailed in

Table 1. Climate change scenarios.

Scenarios	Climate	Land Use
M1	1980s	1985
M2	1990s	1985
M3	1990s	1995
M4	2000s	1995
M5	2000s	2005
M6	2010s	2005
M7	1980s	1985
M8	2010s	1985

. For example, the scenario quantifying the contribution of climate change to water conservation capacity during the 1990s can be presented as follows:

$$R_{1990s}={\displaystyle \frac{{WC}_{M2}-{WC}_{M1}}{\left|{WC}_{M1}\right|}}\times 100\%$$

(4)

where R_1990s is the variation in climate change’s contribution to water conservation capacity between 1991–2000 and 1981–1990, %; WC_M2 is the average annual water conservation during the 1990s, mm, which is calculated under the 1980s’ land use scenarios; WC_M1 is the average annual water conservation for the period 1981–1990, mm; which is also under the land use scenarios of the 1980s. Similar methodologies are applied to obtain other scenarios, as illustrated through calculations M4 with M3, M6 with M5, and M8 with M7. Considering the hydrological cycle’s inherent temporal delay in various months, WC_M1 has negative values; thus, the denominator is expressed in absolute terms here.

The detailed descriptions of the SWAT model construction, and validation, and the Mann–Kendall test methods can be found in Supplementary Materials.

3.4. SWAT Model

3.4.1. Construction of the SWAT Model

The Soil and Water Assessment Tool (SWAT), a watershed-scale model conceptualized by the USDA Agricultural Research Service (ARS), is instrumental in modeling both the qualitative and quantitative aspects of surface runoff and groundwater. It is adept at forecasting the prolonged effects of land management strategies on hydrological dynamics, sedimentation, and the output of agricultural chemicals within extensive and complex watersheds, characterized by varied soil types, land uses, and management scenarios [44]. The foundational theory of SWAT is detailed in Equation (4) [45]. This study focuses on the simulation of hydrological processes, necessitating the input of both spatial and attribute data into the SWAT model. This encompasses land use, soil distribution, and DEM as spatial data, along with meteorological and soil databases as attribute data. The SWAT model initially segments the area into sub-basins using the DEM, subsequently categorizing hydrological response units based on variations in land use, soil properties, and slope. For this study, a delineation threshold of 383,651 ha was established, resulting in the identification of 25 distinct sub-basins, with their precise locations enumerated in Table S1.

$${SW}_t={SW}_0+{\sum }_{i=1}^t\left(R_{day}-Q_{surf}-E_a-Q_{gw}-W_{seep}\right)$$

(5)

where SW_t (mm) is the soil’s final water content, SW₀ (mm) is the soil’s initial water content, R_day (mm) is the precipitation on the ith day, Q_surf (mm) is the surface runoff, E_a (mm) is the actual evaporation, Q_gw (mm) is the return flow, and W_seep (mm) is the percolation and bypass flow exiting the bottom of the soil profile.

3.4.2. Calibration and Validation of SWAT

After the operation of the SWAT model, parameter calibration was executed using SWAT-CUP. This study adopted several standard evaluation metrics, including the Nash–Sutcliffe Efficiency (NSE), the coefficient of determination (R²), percent bias (PBIAS), and root-mean-square error to standard deviation ratio (RSR), to appraise the model’s efficacy. The NSE, ranging from negative infinity to 1, indicates improved model congruence with observed data as it approaches 1. The R² value, varying from 0 to 1, signifies enhanced alignment of the model with observational data as it nears 1 [46]. PBIAS measures the deviation extent between simulation outputs and actual measurements, with smaller values denoting closer alignment. RSR serves as an indicator for assessing the model’s fit and residual attributes, with values approaching zero indicating a greater adherence of the regression model’s residuals to the underlying model assumptions.

$$R^2={\displaystyle \frac{{\sum }_{i=1}^n{\left[\left(Q_{o,i}-Q_{avg}\right)\left(Q_{m,i}-Q_{mavg}\right)\right]}^2}{{\sum }_{i=1}^n{\left(Q_{o,i}-Q_{avg}\right)}^2{\sum }_{i=1}^n{\left(Q_{m,i}-Q_{mavg}\right)}^2}}$$

(6)

$$NSE=1-{\displaystyle \frac{{\sum }_{i=1}^n({Q_{o,i}-Q_{m,i})}^2}{{\sum }_{i=1}^n{\left(Q_{o,i}-Q_{avg}\right)}^2}}$$

(7)

$$PBIAS=100\times {\displaystyle \frac{{\sum }_{i=1}^n(Q_{o,i}-Q_{m,i})}{{\sum }_{i=1}^n{Q}_{o,i}}}$$

(8)

$$RSR=\left[{\displaystyle \frac{\sqrt{{\sum }_{i=1}^n{\left(Q_{o,i}-Q_{m,i}\right)}^2}}{\sqrt{{\sum }_{i=1}^n{\left(Q_{o,i}-Q_{avg}\right)}^2}}}\right]$$

(9)

where Q_o,i is the observed flow, Q_m,i is the simulated flow, Q_avg is the average of the observed flow, and Q_mavg is the average of the simulated flow. Generally, it is considered that 0.75 < NSE ≤ 1.00, 0.00 < RSR ≤ 0.50 and |PBIAS| < 10% indicate very good results; 0.65 < NSE ≤ 0.75, 0.50 < RSR ≤ 0.60 and 10% ≤ |PBIAS| < 15% denote good results; 0.50 < NSE ≤ 0.65, 0.60 < RSR ≤ 0.70 and 15% ≤ |PBIAS| < 25% represent satisfactory results [47].

3.5. Mann–Kendall Test Method

3.5.1. Trend Test

In this research, the Mann–Kendall (MK) trend test is utilized to assess the trends in temperature, precipitation, and water conservation capacities (annual, rainy season, and dry season) within the UMRB over the period 1981 to 2020. The Mann–Kendall trend analysis is a robust statistical technique for identifying trends within time-series datasets, renowned for its non-parametric properties and versatility in data distribution compatibility, thereby comprehensively unveiling data trends. Predominantly applied in hydrological and meteorological data analyses, the methodology for its computation is outlined as follows:

$$S={\sum }_{i=1}^{n-1}{\sum }_{j=i+1}^{n}sgn\left(x_j-x_i\right)$$

(10)

$$sgn\left(x_j-x_i\right)=\left\{\begin{array}{c}1, x_j>x_i\\ 0, x_j=x_i\\ -1, x_j<x_i\end{array}\right.$$

(11)

where x_j, x_i represent the actual observed values for the respective years j and i, with the condition that j > i, and n is the number of observations:

$$Z=\left\{\begin{array}{c}{\displaystyle \frac{S-1}{\sqrt{Var\left(S\right)}}}, S>0\\ 0, S=0\\ {\displaystyle \frac{S+1}{\sqrt{Var\left(S\right)}}}, S<0\end{array}\right.$$

(12)

$$Var\left(S\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)}{18}}$$

(13)

where Z is the statistic of the standard normal distribution, Var(S) is the variance. Z > 0 suggests an ascending trend, whereas Z < 0 implies a descending trend. Z > 1.96 is indicative of a statistically significant uptrend, and conversely, Z < −1.96 signals a statistically significant downtrend.

3.5.2. Change Point Test

The change point test is employed to detect abrupt changes in the time-series data of temperature, precipitation, and water conservation capacities during the annual, rainy, and dry seasons in the UMRB over the period 1981–2020. The methodology for this assessment is defined as follows:

$${UF}_k={\displaystyle \frac{\left(S_k-E\left(S_k\right)\right)}{\sqrt{Var\left(S_k\right)}}} \left(k=\mathrm{1,2},\dots n\right)$$

(14)

$$E\left(S_k\right)={\displaystyle \frac{n\left(n+1\right)}{4}}$$

(15)

$$V_{ar}\left(S_k\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)}{72}}$$

(16)

where UF₁ = 0, E(S_k) and Var(S_k) are the mean and variance in the cumulative data, respectively, UB_k is the reverse sequential values of the time-series, and

$$\left\{\begin{array}{c}{UB}_k=-{UF}_k\\ k=n+1-k\end{array}\right.$$

(17)

In the analysis, the relationship between UF_k and UB_k is scrutinized, UF_k > 0 is an uptrend, and UF_k < 0 is a downtrend. The significance of these trends is indicated when the UF curve surpasses the bounds of the confidence interval. A change point is identified at the juncture where the UF and UB curves intersect within the confines of the confidence interval.

4. Results

4.1. Model Evaluation

In the UMRB, the construction of numerous dams post-2007 significantly influenced river runoff [48]. This study utilized monthly runoff data from 1980 to 2007 for model calibration and validation. The initial period of 1980–1981 is designated as the warm-up period, the calibration period is from 1982 to 2002, and the validation period is from 2003 to 2007. The comparative analysis between the simulated and observed data reveals a close alignment (Figure 2). In addition, the NSE, PBIAS, R², and RSR are 0.92, 6.4%, 0.93, and 0.28 during the calibration period, respectively; and they are 0.92, 2.2%, 0.92, and 0.28 during the validation period, verifying the robustness of the developed SWAT models.

4.2. Temporal and Spatial Variations in Water Conservation Capacity in the UMRB from 1981 to 2020

4.2.1. Annual Variations

Annual water conservation in the UMRB from 1981 to 2020 ranged between 191.6 and 392.9 mm (Figure 3(a1)), with a Z value of −1.41 indicating a non-significant downtrend (Figure 3(a2)). Before 1998, the UF curve exhibits significant fluctuations, ascending above 0 between 1999 and 2009 and descending below 0 post-2010, with both trends being statistically non-significant (Figure 3(a2)). The intersection points of the UF and UB curves for annual water conservation around the years 1988 and 2008 fall within the α = 0.05 significance level, indicating years of notable change in water conservation in the UMRB.

Across the 25 sub-basins (Figure 4a), annual water conservation varies from 42.3 to 602.7 mm, with higher values in the southern regions (Figure 4(b1)). Using the automated breakpoint method in GIS, the water conservation capacities were categorized into medium, good, and excellent, with respective thresholds of 200 mm and 500 mm. Ten sub-basins (1–8, 10, 13) fall into the medium category, while sub-basins 20, 23, 24, and 25 are rated excellent. The Z value spectrum ranges from −2.67 to 0.48 (Figure 4(c1)), with sub-basins 4 and 10 exhibiting Z values above 0, and the other 23 sub-basins displaying values below 0, indicating a prevalent decrease in water conservation. Sub-basin 16, covering Pu’er City, Yunnan, shows a significant decline, suggesting increased flood risks and water scarcity for its 1 million residents.

4.2.2. Variations in Rainy and Dry Seasons

During the rainy season of 1981–2020, water conservation in the UMRB ranged from 218.6–420.3 mm (Figure 3(b1)), with a Z value of −1.62, indicating a non-significant decreasing trend. The UF curve exhibits notable fluctuations before 1998 (most of the time > 0), ascending above 0 between 1999 and 2010, signifying an increasing trend, and descending below 0 post-2011, denoting a decreasing trend, both trends being statistically non-significant (Figure 3(b2)). The intersection points of UF and UB curves around 2010 and 2013 within the α = 0.05 significance level point to abrupt changes in water conservation during these years. From 1981 to 2020, the UMRB’s sub-basins average rainy season water conservation ranged from 63.6 to 614.0 mm (Figure 4(b2)), showcasing a south-high–north-low spatial pattern, classified into low (<200 mm), medium (200–500 mm), and high (>500 mm). Sub-basins numbered 1–8, 10, 11, and 13 are classified as low, while 15, 20, 23, and 25 are classified as high. The Z values vary from −2.41 to 1.13 (Figure 4(c2)), with most sub-basins showing a decline (except sub-basins 10 and 11). Sub-basin number 16 exhibits a marked decrease (same as annual water conservation).

During the dry season from 1981 to 2020, water conservation in the UMRB ranged from −57.0 to 52.2 mm (Figure 3(c1)), with the Z value of −0.36, suggestive of a non-significant decrease. The UF curve for dry season water conservation remained below 0 post-1984, indicating a general downtrend, predominantly non-significant, with notable changes around 1981, 1983, 2015, and 2017 (Figure 3(c2)). From 1981 to 2020, the average dry season water conservation in the sub-basins ranged from −62.2 to 60.4 mm (Figure 4(b3)), classified as below −40, between −40 and 0, and above 0. The sub-basins with positive water conservation are numbered 11, 20, and 24. The range of Z values is from −3.32 to 1.33 (Figure 4(c3)), with 10 sub-basins having negative Z values (6, 8–12, 15, 16, 17, and 19) and 15 sub-basins having positive Z values. Sub-basin 11 shows a significant decrease. Declining water conservation in the dry season is probably exacerbating drought in these areas.

4.3. Unique Contribution of Climate Change to Water Conservation in the UMRB

4.3.1. Annual Contribution

During the 1990s, the climate change contribution rates across the 25 sub-basins range from −19.6% to 49.2%. Sub-basins 3, 15, 17, 19, 20, 23, and 24 register reductions in water conservation due to climate change, with sub-basin 3 experiencing the most significant decrease. The rest witnessed increases, most notably in sub-basin 12 (Figure 5(a1)). In the 2000s, these contribution rates fluctuated between −32.8% and 12.7%, with sub-basins 1, 2, 4, 5, 13, 15, and 19 showing increments, particularly in sub-basin 19; the others showed declines, most substantially in sub-basin 8 (Figure 5(a2)). In the 2010s, the range was −49.8% to 6.8%, with sub-basins 3, 18, 21, 22, 23, and 25 increasing, especially in sub-basin 3; the remaining sub-basins saw reductions, with sub-basin 13 having the most considerable decrease (Figure 5(a3)). Over the 40 years, the rates vary from −44.0% to 4.3%, with increases in sub-basins 1, 2, 4, 5, 21, 22, and 25, the most pronounced being in sub-basin 5, while the others experience negative impacts, the most significant being in sub-basin 13 (Figure 5(a4)). Thus, from 1981 to 2020, climate change most positively affected annual water conservation in various regions of Qinghai and Tibet, China, while the most significant negative effects are observed in several areas of Lincang and Pu’er cities in Yunnan, China. From a basin-wide view, the 1990s saw a 3.9% increase relative to the 1980s, followed by a 6.0% decrease in the 2000s from the 1990s, and a further 6.7% reduction in the 2010s from the 2000s, cumulating in an 8.7% decline during 1981–2020 (Figure 6(a1)). The impact of climate change varies, showing both positive and negative effects, with a total impact always below 9%.

4.3.2. Unique Contribution in Rainy and Dry Seasons

During the rainy season of the 1990s, the contribution rates of climate change to the 25 sub-basins range from −21.7% to 38.4%, with sub-basins 3, 15, 17, 20, 23, and 24 experiencing reductions, most notably in sub-basin 3. Conversely, all other sub-basins witnessed increasing water conservation due to climate change, with the highest increase in sub-basin 12 (Figure 5(b1)). In the 2000s, this range shifted from −32.0% to 10.1%, with sub-basins 1, 2, 3, 4, 5, 11, 15, 19, and 20 showing increases, particularly in sub-basin 11, while others saw decreases, notably in sub-basin 8 (Figure 5(b2)). In the 2010s, the contribution rates varied from −44.6% to 2.6%, with increases in sub-basins 3, 18, 21, 22, and 25, especially in sub-basin 25; others experienced reductions, most significantly in sub-basin 13 (Figure 5(b3)). Over the entire period, the rates fluctuated between −39.3% and 12.3%, with sub-basins 11, 19, and 25 experiencing increases, most notably in sub-basin 11, while climate change’s negative contributions were most pronounced in sub-basin 13 (Figure 5(b4)). Consequently, from 1981 to 2020, climate change had the greatest positive impact on rainy season water conservation in parts of Tibet and Yunnan, China, with the most adverse effects concentrates in specific areas of Lincang and Pu’er cities in Yunnan, China.

In the dry season of the 1990s, climate change’s contributions to the 25 sub-basins fluctuated substantially, ranging from −334.6% to 58.7%. Sub-basins 1–5 and 12 registered increases, with sub-basin 3 experiencing the most significant rise. Contrarily, all other sub-basins observed decreases in water conservation due to climate change, with sub-basin 15 facing the most substantial decline (Figure 5(c1)). During the 2000s, these contributions varied between −218.1% and 76.2%, with sub-basins 6–8, 10, 13, 18, 19, 21–23, and 25 seeing enhancements, particularly in sub-basin 19. Other regions witnessed reductions, most notably in sub-basin 3 (Figure 5(c2)). In the 2010s, contribution rates spanned from −589.3% to 161.1%, with sub-basins 11 and 19 showing the most significant decreases; meanwhile, all other regions experienced increases in water conservation, the largest being in sub-basin 25 (Figure 5(c3)). Over the entire 40-year period, the contribution rates are observed between −259.3% and 168.9%, with sub-basins 1–5, 13, 15, 18, and 20–25 recording increases, the most pronounced in sub-basin 25. All other sub-basins demonstrate negative impacts, with sub-basin 19 exhibiting the greatest decrease (Figure 5(c4)). Consequently, from 1981 to 2020, climate change exerted the most substantial positive influence on dry season water conservation in certain regions of Laos and Myanmar. In contrast, the most adverse impact is observed in southern Jinghong, Xishuangbanna, Yunnan, China.

On a basin-wide scale, during the rainy season from 1981 to 2020, climate change’s unique contribution increased by 6.3% in the 1990s compared to the 1980s, then decreased by 5.2% in the 2000s and further by 8.8% in the 2010s. Over these 40 years, this results in an overall decline of 8.1%, a modest change that consistently stays below 9% (Figure 6(a2)). The observed decrease in rainy season water conservation contributed to heightened flood peak flows and increased negative flood risks in riverine areas. Conversely, in the dry season, there was a 281.8% decline in the 1990s from the 1980s, a 17.0% decrease in the 2000s, and a 59.9% increase in the 2010s, totaling a 65.2% decrease from 1981 to 2020. (Figure 6(a3)). This −65.2% decline in water conservation during the dry season due to climate change over 40 years indicates the significant drought risks, which can potentially cause crop yield reduction and affect the farmers’ livelihoods, threaten hydropower generation and energy supply, disrupt ecosystems, and exacerbate water competition among multiple stakeholders.

5. Discussion

5.1. How Does Climate Change Contribute to Water Conservation Capacity in the UMRB?

From 1981 to 2020, a notable decline in water conservation within the UMRB is observed annually and across both rainy and dry seasons, primarily due to temporal changes in precipitation. First, the changes in water conservation across different seasons closely mirror precipitation trends and diverge from evapotranspiration and runoff patterns (Figure 6(b1–b3)). Second, pivotal years marking abrupt changes in water conservation, specifically in 1988 and 2008 for annual, 2010 and 2013 for rainy seasons, and 1981, 1983, 2015, and 2017 for dry seasons, coincide with major changes in precipitation (

Table 2. Z values and change point years for annual, rainy season, and dry season precipitation, evaporation, runoff, and water conservation.

Factors	Z	Change Point Value
Annual Precipitation	−0.90	1982 1986 1989 2008 2017 2018
Annual Evaporation	1.41	1994
Annual Surface Runoff	−2.06	2002
Annual Water Conservation	−1.41	1988 2008
Rainy Precipitation	−0.78	1982 2014 2016 2018
Rainy Evaporation	2.46	1997
Rainy Surface Runoff	−1.25	2003
Rainy Water Conservation	−1.62	2010 2013
Dry Precipitation	−1.06	1982 1984 1985 1994 2015 2019
Dry Evaporation	0.13	1981 1984 1987 1988 2002 2005 2008 2014
Dry Surface Runoff	−1.88	1987 1995 1996
Dry Water Conservation	−0.36	1981 1983 2015 2017

). Third, both precipitation and water conservation show a decreasing trend in their Z values (below 0) (Table 2). For specific decades, in the 1990s and 2010s, the annual and rainy season water conservation (Figure 7(a1,b1)) alongside the dry season for the 2000s and 2010s (Figure 7(c1)) indicate that actual evaporation and surface runoff changes align with those of water conservation, ruling them out as change drivers in these periods, so the water conservation during these periods is dominated by precipitation. Therefore, precipitation emerges as the predominant variable, as increased precipitation boosts soil moisture, vegetation growth, and soil water retention, enhancing water conservation [49]. Conversely, decreased precipitation reduces soil moisture and groundwater levels, impairing water conservation and heightening drought risks [50].

Actual evaporation plays a key role in water conservation capacity. In the 2000s, there was a notable 16.5 mm reduction (−5.3%) in rainy season water conservation, coinciding with a 17.4 mm surge in evaporation (4.3%) (Figure 7(b1,b2)). Also, the 1990s experienced a 7.1 mm decrease (−281.8%) in dry season water conservation, alongside a 5.43 mm rise in evaporation (3.8%) (Figure 7(c1,c2)). These trends indicate that evaporation influences water conservation. Rising temperatures accelerate evaporation in water bodies, leading to faster moisture loss from lakes, rivers, and soil, which increases atmospheric moisture but reduces surface and soil water volumes [51]. This phenomenon results in a diminution of water volume both at the surface and within the soil, as the evaporated moisture is not promptly compensated by precipitation [52]. The ability to conserve water depends on moisture infiltration, which decreases as evaporation increases, affecting water conservation [53].

Figure 7(a1–c1) shows that trends in surface runoff and water conservation follow parallel trajectories within each decade, indicating that runoff alterations do not drive variations in water conservation. An increase in surface runoff signifies reduced water availability for groundwater recharge and soil moisture, thereby diminishing the water conservation potential [54]. Regions with hard or steep surfaces face greater challenges in water conservation due to increased runoff [55].

5.2. Precipitation’s Dominance in Water Conservation: Temporal and Spatial Analysis

Over the entire 40-year analysis, there is a 25.8 mm decline in annual water conservation (−8.7%), a 25.9 mm decrease in precipitation (−2.8%), and a 14.1 mm increase in actual evaporation (2.6%) (Figure 7(a1,a2)). Also, in the 40 years, there is a 24.1 mm decline in rainy season water conservation (−8.1%), a 19.4 mm decrease in precipitation (−2.5%), and a 10.5 mm increase in evaporation (2.6%) (Figure 7(b1,b2)). This dry season sees a 1.7 mm reduction in water conservation (−65.2%), a 6.5 mm decrease in precipitation (−4.2%), and a 3.7 mm rise in evaporation (2.6%) (Figure 7(c1,c2)). These trends indicate that water conservation changes are primarily driven by precipitation, with evaporation playing a smaller role.

The sub-basin analysis from the 1990s, 2000s, and 2010s shows that over 76% of sub-basins are impacted by precipitation annually (Figure 8(a1–a3)), over 72% during rainy seasons (Figure 8(b1–b3)), and more than 68% in dry seasons (Figure 8(c1–c3)). Over the 40 years, 64% of the UMRB’s sub-basins see annual water conservation mainly influenced by precipitation, while evaporation affects 36% (Figure 8(a4)). In the rainy season, precipitation impacts 48% of sub-basins, and evaporation 40% of sub-basins (Figure 8(b4)). The dry season has a stronger precipitation influence, affecting 92% of sub-basins, compared to evaporation’s 8% impact (Figure 8(c4)). Thus, a comprehensive analysis across annual, rainy, and dry seasons reveals a predominant influence of precipitation on water conservation across the majority of sub-basins.

Thus, precipitation plays the dominant role in modulating water conservation across the UMRB, and its variability directly dictates water resource availability and water conservation capacity, which aligns with previous research [9,53,56,57].

5.3. Assessing the Difference across Annual, Rainy, and Dry Seasons

The influence of climate change rates on water conservation is modest annually and during the rainy season (8.7% and 8.1%, as we mentioned in Section 4.3), whereas its impact is markedly pronounced during the dry season (65.2%, as we mentioned in Section 4.3). The values of the variation in water conservation in the rainy season are greater than those in the dry season (Figure 7(b1,c1)). Due to the higher baseline of water conservation in the rainy season, the rate of change appears lower. Since annual water conservation largely depends on the rainy season, the overall annual impact of climate change remains modest. From 1981 to 2020, over 56% of the sub-basins in the UMRB had contribution rates exceeding 10% throughout the year, including both rainy and dry seasons (Table S2). The effects of climate change on the UMRB are significant, with positive and negative impacts in different regions that balance out to produce minor variations across the basin. Further, an intra-annual analysis shows substantial fluctuations during the rainy season (May–October) from 1981 to 2020 (Table S3). July and August exhibit positive rates, while other months show negative contributions, reducing the overall impact on the rainy season. The dry season displays both positive and negative monthly variations, with relatively significant variation rates (Table S4). Consequently, the UMRB has been adversely affected by climate change, yet the complexities of geographical conditions and uneven intra-annual distribution culminate in a minor net annual contribution rate.

5.4. Developing Intervention Strategies for Augmenting Water Conservation Capacity in the UMRB

Addressing the unique challenges in the Qinghai–Tibet and Yunnan–Myanmar sections of the UMRB necessitates specific strategic interventions. Primarily, in the Qinghai–Tibet high-altitude region, the focus should be on reinforcing the protection and rejuvenation of alpine grasslands. Given their exceptional capacity for water conservation, safeguarding these grasslands is imperative for curtailing water loss. Key initiatives include preventing overgrazing and deforestation, thereby preserving the ecological integrity of these grasslands. In contrast, the Yunnan and Myanmar segments, characterized by a mix of mountainous and hilly terrains, demand a segmented approach to water resource management. Forest conservation in mountainous areas should be prioritized to bolster water conservation, while in hilly regions, the promotion of integrated agroforestry practices and curtailing the expansion of agricultural lands are essential to mitigate soil erosion. Tailoring strategies to suit the diverse topographies ensures an optimized enhancement of water conservation capabilities across the UMRB. Furthermore, in light of the profound implications of climate change on water resources, the adaptive strategies are indispensable, such as the establishment of robust early warning systems, the extreme climatic events management, and the long-term climate adaptation policy frameworks formulation, etc.

6. Conclusions

This study used meteorological, DEM, land use, and soil data to construct the SWAT models, coupled with the Mann–Kendall test method, to intricately investigate the water conservation capacity dynamics in the UMRB from 1981 to 2020. The major conclusions are as follows.

(1) The annual water conservation capacity fluctuates between 191.6 and 392.9 mm, with higher values in the rainy season (218.6–420.3 mm) than that in the dry season (−57.0 to 53.2 mm). Both rainy and dry seasons demonstrate a non-significant declining trend. Additionally, spatial heterogeneity in annual water conservation capacities is observed across different sub-basins, with higher capacities in the southern regions and lower capacities in the northern regions, and a general decreasing trend in most sub-basins.

(2) The contribution of climate change to water conservation capacities was −8.7% annually, −8.1% during the rainy season, and −65.2% in the dry season over the past 40 years from 1981 to 2020. The contributions of climate change exhibit significant spatial variations among different sub-basins, with sub-basin 25 the most positive impact and sub-basin 13 the most negative impact.

(3) The influences of precipitation, actual evaporation, and surface runoff on water conservation capacity presented significant temporal differences. Across the 40-year analysis, precipitation emerges as the primary factor in all seasonal variations in water conservation capacity. Consequently, the precipitation patterns change caused by climate change are identified as the most significant factor affecting water conservation capacity, followed by evaporation and runoff.