SB: Climate change will undermine the pivotal role of Asian water tower in water resources supply

The Tibetan Plateau and surrounding Hindu Kush, Karakoram, and Himalayan mountain ranges, often referred to as the “Asian water tower (AWT)” [1], is a critical hydrological system that is under significant threat due to climate change. The AWT is home to the largest reserves of ice outside the Arctic and Antarctic [2]. It is the birthplace of many Asian rivers such as the Indus, Ganges, Brahmaputra, Yangtze, Yellow, Mekong, Salween, Tarim, Amu Darya, and Syr Darya (Table S1 online), providing important water resources for the downstream plains [[3], [4], [5]]. However, climate change is gradually changing the hydrological cycle in this region [[6], [7], [8]] and has far-reaching implications for hydropower, agriculture, and ecosystems at riparian countries [9,10].

Glaciers in the AWT have been retreating at an alarming rate due to global warming [[11], [12], [13]]. The reduction in glacier volume and area has direct implications for the water supply to downstream regions. Glaciers act as natural reservoirs, storing water during the colder months and releasing it during the warmer months. As glaciers retreat, this storage capacity decreases, leading to a reduction in the availability of water during the dry season. This will finally lead to a significant reduction in the hydrological contribution of mountainous regions to downstream water resources. Second, the AWT is one of the regions that are most sensitive to climate change, with warming rates approximately twice the global average [2]. This temperature rise may lead to increased evapotranspiration rates in the mountainous regions, which in turn reduces the amount of water available for streamflow. The increased temperature not only accelerates the melting of glaciers but also increases the rate at which water is lost to the atmosphere through evapotranspiration. Third, the increasing rates of precipitation of the mountainous AWT [2] can be different from those in the downstream plains, which will influence the relative contributions of mountain runoff to downstream freshwater supply.

Changes in the hydrological contributions of the AWT may have significant implications for downstream water resources at various aspects [[2], [3], [4],9], e.g., (a) water security and availability (agriculture, industry, and domestic water use), (b) agriculture and food security, (c) hydropower generation and energy security, and (d) ecosystem services and biodiversity. Despite the prominent importance of the AWT’s mountain runoff for the downstream human society and natural ecosystem, the future hydrological contributions of AWT to downstream water resources are still unknown under climate change in the 21st century.

This paper aims to quantify the future changes in the hydrological contributions of the AWT to downstream water resources and uncover the potential driving forces at ten large Asian rivers, including the Indus, Ganges, Brahmaputra, Yangtze, Yellow, Mekong, Salween, Tarim, Amu Darya, and Syr Darya. Three primary factors, including the glacier retreat, the elevation-dependent warming and evapotranspiration trends, and the spatial variations in precipitation changes, will be examined in detail. Finally, the implications and adaptation measures will be discussed regarding the downstream water security.

The projections of future natural runoff at the mountain basins and the whole basins (Supplementary methods online) have been performed at each of the ten major rivers originating from the AWT. Results show that, the mountain-basin runoffs at most of the studied rivers show significant increasing trends (Fig. S5a, c online) that are generally consistent with the rising mountain precipitation (Fig. S7a, c and Table S2 online), except for the Amu Darya and Syr Darya rivers that have opposite trends under both a moderate warming scenario (SSP245) and a high warming scenario (SSP585). Compared to the mountain-basin runoffs, the whole-basin runoffs at the seven studied rivers (the Ganges, Brahmaputra, Salween, Yellow, Tarim, Mekong and Yangtze rivers) tend to have larger rates of increasing trends under both SSP245 and SSP585 (Fig. S5 and Table S2 online), except for the Syr Darya river (decreasing trends), the Amu Darya river (no significant trends) and the Indus river (no significant trend under SSP245, but lower increasing rate under SSP585).

Under the moderate warming scenario (SSP245), 7 out of the 10 studied rivers exhibit significantly declining trends in CR (Fig. 1b), due to the higher growth rates of future whole-basin runoffs than those of mountain runoffs (Fig. S5a, b and Table S2 online). The Syr Darya river shows an opposite CR trend (significantly increasing), while the Indus and Yellow rivers have no significant CR trends under SSP245 (Fig. 1b, Table S3 online). The significantly positive CR trend in the Syr Darya river can be attributed to the slower decreasing rates of the mountain runoffs (–1.1%/decade) compared to that of the whole-basin runoffs (–1.5%/decade) (Table S2 online). The insignificant CR trend at the Indus river or the Yellow River is generally a result of the comparable changing rates between the mountain runoffs and the whole-basin runoffs under SSP245 (Fig. S5a, b and Table S2 online).

From the present to the near future, almost all of the studied rivers show declines in their dependence on the mountain runoffs except for the Indus river, under both SSP245 and SSP585 (Fig. 2c). Among all the studied rivers, the Tarim river shows the most significant decline in CR from 73.3% at the present to 48.9% ± 0.6% in the near future under SSP245 (51.9% ± 0.7% under SSP585; Fig. 2c), due to much larger increase of natural runoffs in the whole basin (56.8% ± 1.7% under SSP245, and 50.0% ± 2.1% under SSP585) than that in the mountain basins (4.5% ± 1.4% under SSP245, and 6.2% ± 1.5% under SSP585) (Fig. 2a). Similarly, the Yellow River shows the second-largest decline in CR, from 36.3% at present to 23.5% ± 1.4% in the near future under the SSP245 scenario (22.7% ± 1.2% under SSP585; Fig. 2c), due to much larger increase of natural runoffs in the whole basins (79.4% ± 5.5% under SSP245, and 89.0% ± 6.1% under SSP585) than that in the mountain basins (16.1% ± 6.9% under SSP245, and 18.1% ± 6.4% under SSP585) (Fig. 2a). Different from the other studied rivers, the Indus river is the only one that becomes more dependent on the mountain runoffs in the near future, showing an increase of CR from 33.6% at the present to 43.2% ± 2.0% in the near future under SSP245 (37.9% ± 1.4% under SSP585; Fig. 2c), due to both an increase in the mountain-basin natural runoffs (5.1% ± 4.8% under SSP245, and 6.8% ± 3.8% under SSP585) and a decline in the whole-basin natural runoffs (–18.3% ± 1.6% under SSP245, and –5.4% ± 1.0% under SSP585) (Fig. 2a).

From the present to the far future, under both SSP245 and SSP585, most of the studied large rivers also show declines in their dependence on the mountain runoffs except for the Indus river (Fig. 2c). The Indus river demonstrates an increase of CR from 33.6% at the present to 38.5% ± 2.2% in the far future under SSP245 (37.3% ± 1.8% under SSP585; Fig. 2c), which is attributed to an increase (9.7% ± 6.3%) in the mountain-basin natural runoffs and a decline (–4.4% ± 1.5%) in the whole-basin natural runoffs under SSP245, and much larger increases in the mountain-basin natural runoffs (16.6% ± 5.7%) than that in the whole-basin natural runoffs (4.9% ± 2.4%) under SSP585 (Fig. 2b). Among all the studied large rivers, the Tarim river again shows the most significant decline of CR from 73.3% at the present to 48.6% ± 0.8% in the far future under SSP245 (52.7% ± 0.9% under SSP585; Fig. 2c), due to much larger increases of the natural runoffs in the whole basins (60.1% ± 3.4% under SSP245, and 60.3% ± 2.8% under SSP585) than that in the mountain basins (6.1% ± 1.8% under SSP245, and 15.2% ± 1.9% under SSP585) (Fig. 2b).

Driving forces behind the declining contributions of the Asian water tower to the downstream have been discussed below from three aspects.

Firstly, future precipitation changes are projected to have distinct spatial patterns, with different implications for mountainous and downstream areas. The Inter-Sectoral Impact Model Intercomparison Project climate models (Supplementary datasets online) from the World Climate Research Program (WCRP) Coupled Model Intercomparison Project Phase 6 (CMIP6) suggest that precipitation in all the ten studied rivers will increase for both the upstream and downstream basins, but not uniformly (Fig. S7 and Table S2 online). In the northern arid AWT basins (e.g., the Amu Darya, Syr Darya, Yellow, and Tarim basins), precipitation is expected to increase less than that in the downstream plains (Fig. S8 online). In contrast, the southern wet AWT basins (e.g., the Brahmaputra, Mekong, and Yangtze basins), are expected to experience greater precipitation increases in the mountains compared to the plains (Fig. S8 online). Secondly, elevation-dependent warming and evapotranspiration (ET) increases may lead to a greater rate in runoff decline at the mountain basins compared to downstream lowlands. Temperatures in the high mountain regions of the AWT have been increasing at a faster rate than the global average [2,14]. At most of the studied rivers, the mountain warming rates are greater than those in the downstream plains under both SSP245 and SSP585 (Fig. S9 online). Compared to downstream lowlands, the greater warming rates at higher altitudes have led to a faster rise in ET rates (Figs. S10 and S11 online). As more water is evaporated from the soil and transpired by plants, less water is available to contribute to river flow, leading to a decrease in mountain runoff. Thirdly, vanishing glaciers in the AWT may ultimately reduce the mountain runoffs, leading to declining contributions of mountain runoff to downstream water resources by the end of the 21st century. Glaciers in the AWT have been retreating at an alarming rate due to global warming [2,[11], [12], [13]], and the reduction in glacier volume and area has direct implications for the water supply to downstream regions. Results show that the regional glacier mass-loss rates in all the studied mountain basins generally increase at first until reaching a peak and then decrease, under both SSP245 and SSP585 in the 21st century (Fig. S12 online). As a result, the glacier meltwater (represented by the excess discharge of glaciers; Supplementary methods online) will also experience an increase in the near future and a final decline in the far future (Figs. S13 and S14 online). This will lead to a significant reduction in the hydrological contribution of mountainous AWT to downstream water resources in the far future, particularly for the Brahmaputra, Salween, Yangtze, Syr Darya, Ganges, and Mekong rivers (Fig. S14 online). In contrast, due to the large glacier storage, the glacier melt at the Indus, Tarim, and Amu Darya rivers will not experience a marked decline even in the far future compared to the present (Fig. S14 online). Therefore, the spatial patterns of precipitation, ET, and glacier melt changes have significant implications for river runoff. In the northern arid AWT basins where mountain precipitation is projected to increase less than in the plains, the overall contribution of mountain runoff to downstream water resources may decrease. Conversely, in the southern wetter AWT basins where mountain precipitation is projected to increase more than in the plains, the contribution of mountain runoff may increase, but this increase may be offset by the reduction in glacier meltwater and the increase in evapotranspiration.

In summary, this study examines the impacts of climate change on the hydrological contribution of the AWT's mountainous regions to downstream plains. Major findings from this study are as follows. (1) The hydrological contribution of the AWT to downstream water resources is under threat due to climate change. (2) The retreat of glaciers, increased evapotranspiration, and spatial variability in precipitation increases are the primary reasons for this reduction. These factors, when combined, lead to a decrease in the availability of water from the mountainous regions (except for the Indus river), affecting the water supply to downstream areas.

On the one hand, we should emphasize the marked advantages of the data-driven Long Short-Term Memory (LSTM) model [16] over complex physical models in water resource projection in the AWT region.

First, LSTM ensures high accuracy while being more computationally efficient. Unlike physical models that often require intensive mathematical calculations and parameter tuning, LSTM leverages neural networks to learn patterns from data, reducing computational costs and accelerating projection processes. Second, LSTM demands fewer input data. Physical models typically rely on extensive domain-specific knowledge, detailed physical laws, and large volumes of high-precision data (e.g., meteorological, topographical parameters). In contrast, LSTM can capture underlying relationships in data with relatively limited inputs, making it more adaptable to regions where data availability is constrained. These characteristics make LSTM particularly suitable for water resource projection in the AWT basins, where complex terrain and data scarcity pose challenges for traditional physical models. LSTM’s efficiency and data parsimony enhance its practicality for regional-scale hydrological forecasting.

On the other hand, to complete the picture of future AWT runoff changes and their potential impacts on the downstream water availability, the hydrological impacts of permafrost and groundwater changes, as well as the uncertainty of datasets used in this study, should be further discussed.

Firstly, permafrost changes in the AWT region will have diverse impacts on future runoff [8,15]. In the short term, the thawing of permafrost will release a large amount of stored water, increasing the runoff volume. However, in the long term, as the permafrost layer thins and its water-retaining capacity declines, the runoff may gradually decrease. Secondly, changes in groundwater levels have direct impacts on baseflow and seasonal runoff contributions in the AWT basins and their downstream, which should be accounted for in the next studies. When the groundwater level rises, more groundwater can discharge into rivers, increasing the baseflow. Conversely, a decline in the groundwater level reduces the baseflow, making rivers more prone to drying up in the dry season. Thirdly, we should acknowledge the uncertainties in the multi-source datasets used in this study. (a) The application of ERA5-Land reanalysis data (precipitation and air temperature) into the AWT region has both strengths and limitations. For precipitation, the ERA5-Land can capture the general spatial and temporal patterns to some extent, but has difficulties in accurately representing extreme events and fine-scale variations in complex terrains of the AWT region. For air temperature, the ERA5-Land generally performs better in reflecting the overall temperature trends, yet may have biases in high-altitude and snow-covered areas due to issues in surface energy balance representation. (b) The future climate projections used in this study may have different uncertainties [17,18] under SSP245 and SSP585. (c) The glacier mass balance projections used in this study may contain considerable uncertainties in the far future, which may influence the future mountain runoff projections (e.g., at the upper Indus). Given the challenges posed by climate change, adaptive management strategies are essential to safeguard water resources in the AWT and the downstream plains. (1) Investing in water storage infrastructure, such as reservoirs and dams, can help buffer against the impacts of decreased runoff. This can provide a means of storing water during periods of high flows and releasing it during periods of low flows. (2) Enhancing water use efficiency in agriculture, industry, and domestic use can help reduce the demand for water and make the best use of the available resources. (3) Transboundary cooperation and effective governance are crucial for managing water resources in the Asian rivers, which include sharing data, developing joint adaptation strategies, and managing conflicts over water resources.