1. Introduction
The Tibetan Plateau (TP) is known as the “Asian Water Tower” because it feeds 12 great Asian rivers, including the Yangtze, Yellow, Indus, Mekong, Salween, Ganges, Yarlung Zangbo, Amu Darya, Syr Darya, and Tarim. These rivers provide a crucial and reliable water supply to billions of people (Immerzeel et al. 2010). At the most fundamental level, the concept of a “water tower” refers to components of inflow, storage, and outflow (Immerzeel et al. 2020), which are closely linked to the hydrological process of recharge–storage–runoff. Due to unique topography and specular climate regime over the TP, the inflow consists primarily of precipitation in the form of snow and rain derived from the westerlies and monsoon systems. Redistribution of water surplus refers to flowing into terrestrial water aquifers and rivers. There are widespread glaciers and snow-covered mountain ranges, thousands of lakes, and large areas of permafrost, which delay the release of water and alert the water cycle over the TP.
The recharge–storage–runoff process provides a consistent picture of multiple aspects of the water cycle in a hydrologic unit (Syed et al. 2009). The primary processes include precipitation (P) and evapotranspiration (ET), terrestrial water storage and flow of water on and below the surface, and atmospheric transport of water vapor over land. Recharge is determined by the net water flux (especially P − ET > 0), which controls both storage and runoff at watershed. Terrestrial water storage, consisting of all forms of water stored above and underneath the land surface, serves as an intermediate buffer related to both recharge and runoff (Kuppel et al. 2017; Scanlon et al. 2019). The buffering role is more apparent in the regions covered by snow and ice. Despite having the largest global store of frozen water after the polar region, the buffering role remains elusive over the TP and its basins.
Runoff is the result of a series of discontinuous processes in precipitation, soils, and hillslopes, which are integrated into the recharge–storage–runoff hysteresis within a river basin (Sproles et al. 2015; Riegger and Tourian 2014). Although river discharge naturally integrates a host of water cycle processes over the drainage area, gauge-based measurements present significant challenges worldwide (Dai 2016). The estimation of runoff was analyzed in global and regional scales based on water budget closure (Reeves Eyre and Zeng 2021). Runoff as the net surface and groundwater outflow from a watershed may differ considerably, while it closely relates to river discharge in many regions of the world.
The recharge–storage–runoff process is a key to understanding how a hydrological system is functioning. It represents spatial and temporal dynamics of freshwater exchange at the land margin. The relationship of recharge, storage, and runoff is nonlinear, heterogenetic, and complex. For example, snowfall and rainfall have distinct resident times in the terrestrial water system (Riegger and Tourian 2014). A column with a large increase in water storage in response to the recharge phase can be simply characterized as having high runoff-buffering potential (McNamara et al. 2011). However, the buffering role is impacted by hydrogeologic features and climatic conditions (Scanlon et al. 2019). The climate monitoring network is rather sparse over the TP (Wang et al. 2021). Estimation of areal averages of the recharge–storage–runoff depends on self-consistent combinations of datasets that are a central theme of challenges to be addressed in this study.
The TP, known as a vital and vulnerable water tower, will be impacted profoundly by climatic and socioeconomic changes (Immerzeel et al. 2020; Yao et al. 2022). Although water supply and demand are often linked to river discharge, the buffering role is essential but less known over the TP where multiple topographically different mountain ranges, lakes and glaciers are widely distributed. TP has experienced a warming trend in twice of the global mean rate during the last five decades (Chen et al. 2015). The projection studies infer to a continuous warming trend in the future (Duan and Xiao 2015). Snow– and ice–albedo feedback related to the loss of highly reflective snow and ice cover lead to elevation-dependent warming, though the reported warming rate at very high elevations are controversial (You et al. 2020; Ouyang et al. 2019). Corresponding to the warming, there is a wetting trend with increases in atmospheric moisture, precipitation, and evapotranspiration. The glaciers in the most of the TP have experienced mass loss since the 1970s (Yao et al. 2019, 2012; Li et al. 2022). Major lakes in the central TP have expanded in area and depth since the middle of 1990s (Lei et al. 2017; Zhang et al. 2013, 2017).
Accurately understanding changes in the recharge–storage–runoff process over the TP depends on reliable and comprehensive data sources. This study aims to evaluate the utility of remote sensing and reanalysis datasets for characterizing the recharge–storage–runoff process over the TP and to assess their potential to contribute to a better understanding of water resources. Because of a hypothesis of closure achieved with ERA5 precipitation and evapotranspiration plus Gravity Recovery and Climate Experiment (GRACE) satellite terrestrial water storage changes, runoff is constrained by land–atmosphere water balance and compared with available river discharge measurements. This study seeks to answer several scientific questions regarding the recharge–storage–runoff process over the TP: 1) What is the annual recharge volume over the TP? 2) How is the recharge redistributed into the storage and runoff? The buffering role and its unique characteristics at the basin scale of the TP are particularly interesting.
2. Data and method
a. Terrestrial water balance over the TP
Study areas over the TP (with height greater than 2 km) are shown in Fig. 1. According to great rivers and watersheds provided by the HydroSHEDS dataset, the TP is divided into 12 drainage basins (Amu Darya, Indus, Ganges, Brahmaputra, Salween, Mekong, Yangtze, Yellow, Hexi Corridor, Qaidam, Tarim, and Inner Plateau) (Immerzeel and Bierkens 2012; Zhang et al. 2013). Because river discharge records are available only at the outlet hydrological stations Nignii, Yancun, Tangnaihai, and Zhimenda, four corresponding drainage basins (colored in Fig. 1) are focused on in the investigation. In addition, the Inner basin is studied as the only endorheic basin.
Map of the studied domain over the TP, which is divided into 12 drainage basins (from 1 to 12 named as Amu Darya, Indus, Ganges, Brahmaputra, Salween, Mekong, Yangtze, Yellow, Hexi, Qaidam, Tarim, Inner, respectively). Four discharge stations (Nignii, Yancun, Zhimenda, and Tangnaihai) are marked by crosses, together with corresponding watersheds [that are part of the basins of Amu Darya (1), Brahmaputra (4), Yangtze (7), and Yellow (8)]. The Inner (12) is the only endorheic basin.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
b. Recharge and the net water flux
Recharge is calculated by the net water flux (P − ET > 0) using ECMWF’s latest atmospheric reanalysis dataset ERA5 in the period 1979–2021. ERA5 has a resolution of 31 km and 137 levels to 0.01 hPa, as well as updated model physics and improved data assimilation process (Hersbach et al. 2020). Prognostic and diagnostic methods are used to calculate the net water flux. In the prognostic method, P and ET are taken from numerical forecasts in ERA5. The P and ET are only available from short-term forecasts and are taken from twice-daily 12-hourly forecasts started at 0600 and 1800 UTC. In the diagnostic method, the net water flux is computed by the vertically integrated horizontal moisture flux divergence (VIWVD) and the vertically integrated moisture tendency (QTEND) according to atmospheric water balance. VIWVD and QTEND are computed using analyzed fields. A previous study found that ERA5 in principle is suited for the quantification of atmospheric budgets over the TP, but imbalance is obvious in basins Qaidam and Inner (Lei et al. 2021). Biases in reanalyses are introduced in the region where observational information is lacking.
The diagnostic net water flux (DNWF) used in this study has applied a mass-consistent wind adjustment and a barotropic correction (Mayer et al. 2021). Consistency of the mass budget is achieved by iteratively adjusting the horizontal wind field according to the mass budget residual where parameterized P and ET fluxes are approximated by the analyzed divergence and tendency of the vertically integrated water vapor content. Previous studies found that the difference between mass-adjusted and unadjusted moisture divergence may become larger due to a great reducuction in artificial noise of wind over high topography (Trenberth 1991; Mayer and Haimberger 2012). Differences in the net water fluxes derived from prognostic and diagnostic methods are estimated over the TP.
c. Water storage changes and buffer capacity
Satellite-based estimates of water storage change (dS/dt) are calculated from GRACE (Syed et al. 2005). GRACE and GRACE Follow-On missions launched by NASA and the German Aerospace Center observe time-variable gravity field that is able to provide the total water storage anomalies (TWSA) on a global scale since 2002 (Tapley et al. 2019; Rodell and Reager 2023). Daily TWSA during 2002–19 from ITSG-Grace2018 (Institute of Geodesy at Graz University of Technology; Eicker et al. 2020) is employed to derive the change in terrestrial water storage over the monthly time span [dS/dt of Eq. (1)]. Validation of ITSG-Grace2018 with ERA5 and ERA-Interim had shown that GRACE-derived water flux produced realistic variations on seasonal and interannual time scales (Lei et al. 2021; Eicker et al. 2020).
The dS/dt within large river basins is critical in providing information on storage dynamics, with positive (negative) values for the filling (releasing) of storage water (Wu et al. 2021). This perspective is useful in understanding the seasonal recharge–discharge cycle and the buffer capacity of the watershed basin. Previous study (Kuppel et al. 2017) quantified buffer capacity by averaging the accumulated positive (filling) and negative (releasing) monthly water storage change (dS/dt) over a 12-month period. This study employs a similar approach, but buffer capacity is quantified within the recharge–storage–runoff chain.
d. Runoff and river discharge
Runoff [R of Eq. (1)] is estimated using the terrestrial water balance approach, with the assumption of closure. R represents the balance of surface, groundwater and tidal inflows and outflows (Syed et al. 2009). To measure the fast and slow runoff responses to the net water flux, the ratio (r1) of runoff outflow (R > 0) to the net water flux in the storage filling condition (dS/dt > 0) and the ratio (r2, absolute value) of the storage release (dS/dt < 0) to runoff outflow are calculated as proxies. Small proportions of r1 and large contributions of r2 suggest high runoff-buffering potential, and vice versa.
Ground-based river discharge (Rd) monitor provides important information about the in-channel component, which excludes groundwater discharge and surface flow in braided channels or in inundated floodplains. Thus, Rd may only represent a part of the net freshwater flux in reality. Although Rd is incomplete for budget analyses, the observed rates (mm month−1) at four stations (Nignii in Amu Darya, Yancun in Brahmaputra, Zhimenda in Yangtze, and Tangnaihai in Yellow basin) provide realistic and consistent reference values of R in the terrestrial water balance. These observations have been available since 1979 and have lasted until 2014 or later.
3. Results
The time series analysis of recharge–storage–runoff is conducted for the TP, Inner basin, and subbasins of Amu Darya, Brahmaputra, Yangtze, and Yellow. Figure 2 displays the monthly means of the prognostic and diagnostic net water flux derived from ERA5 during 1979–2021, together with dS/dt from GRACE and GRACE Follow-On, and available river discharge observations. The annual cycle is clearly evident. Net water fluxes over the TP vary from −7 to over 100 mm month−1 throughout the year, with an average of around 30 mm month−1, which suggests freshwater recharge in general. Less than 8% of records are negative over the TP, but more negative net water fluxes occur especially at the Inner basin (50%), subbasins of Brahmaputra (40%) and Amu Darya (30%). Both atmospheric water supply and demand are significant processes at the basin scales. The seasonal imbalance between water surplus and deficit is driven by climatic factors. Nevertheless, water surplus (P − ET > 0) and redistribution in the water storage and runoff are of particular interest.
Monthly mean of prognostic (P − ET in black lines) and diagnostic net water fluxes (DNWF in red) during 1979–2021 over (a) the TP, (b) Inner basin, and (c)–(f) subbasins of Amu Darya, Brahmaputra, Yangtze, and Yellow, respectively. The dS/dt from GRACE and GRACE Follow-On is shown in blue. River discharge observations at Nignii, Yancun, Zhimenda, and Tangnaihai are illustrated in green (Rd). Units are mm month−1 on the left axes. Using a 11-month running mean, smoothed temporal evolutions of P − ET, DNWF, dS/dt, and Rd are displayed in the bottom with the right figure axes.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The variations of dS/dt show a phase relationship with the net water flux, with correlation coefficients of 0.46 over the TP, 0.58 over Inner, 0.9 over Amu Darya, 0.66 over Brahmaputra, 0.62 over Yangtze, and 0.71 over Yellow. The annual means of dS/dt at the TP and basin scales are close to zero. The seasonality of dS/dt suggests that precipitation is stored and subsequently released, indicating hydrological buffer capacity (Kuppel et al. 2017). Water storage (dS/dt > 0) and release (dS/dt < 0) occur throughout the year, with different timing and amount representing the inherent complexity of watersheds at basins and broader TP scales. The investigation of the recharge–storage–runoff process-based understanding of the seasonal water buffering characteristics and drivers of water cycle is emphasized.
The river discharge at Amu Darya is found to be out of phase with the net water flux and dS/dt, with correlation coefficients of −0.66 and −0.78, respectively. This highlights the essential buffering role played by the Amu Darya. On the other hand, river discharge is in phase with the net water flux at subbasins of Brahmaputra (0.76), Yangtze (0.69), and Yellow (0.67), suggesting fast responses to water surplus. The seasonality in dS/dt implies a disparate pattern of the buffering role associated with snowfall or rainfall, which will be discussed later.
a. Comparison of the prognostic and diagnostic net water fluxes
To accurately assess the recharge–storage–runoff over the TP and basins, it is critical to have self-consistent combinations of data. This study assumes water budget closure, and the accuracy of the net water flux is crucial in providing valuable constraints to budgets of storage and runoff. In Fig. 2, 11-month running means highlight that these is an overestimation of the prognostic net water flux (the average of 34 mm month−1) compared to the diagnostic net water flux (27 mm month−1) over the TP. The annual mean of the diagnostic net water flux is 326 mm yr−1, which is a 21% decrease compared to the prognostic net water flux of 412 mm yr−1. Based on the prognostic net water flux accounting for 51.7% of the annual precipitation in ERA5, the DNWF constrains precipitation with 629 mm yr−1 over the TP.
Regional characteristics in differences between the prognostic and diagnostic net water fluxes are evident, though there is a close relationship between P − ET and DNWF (correlations ranging from 0.96 to 1). Figure 3 displays the spatial distribution of difference in the net water fluxes. The edge (interior) of the TP tends to overestimate (underestimate) P − ET in compared to DNWF. The diagnostic net water flux significantly reduces artificial noise over the high topography (Mayer et al. 2021). Previous studies suggested that this improvement is mainly due to advection rather than wind divergence (Trenberth 1991; Mayer and Haimberger 2012), which partially explains the spatial pattern in Fig. 3. The high topography of the TP reduces advection of wind on the edge, while the terrain effect in the interior decreases.
Difference of the prognostic net water flux (P − ET; mm month−1) with the diagnostic (DNWF) during 1979–2021.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
Figure 4 presents ratios (the median, minimum and maximum, and 25th and 75th percentiles) of P − ET to the diagnostic net water flux, focusing only on water surplus (DNWF > 0), over the TP, Inner and four subbasins. The results show that overestimations are prevalent, with median values of 1.42 and 1.26 at the TP and subbasin of Brahmaputra, respectively. Conversely, underestimations are visible at the Inner basin (0.88), while the median ratio is approximately 1 at subbasins of Amu Darya (1.01), Yangtze (0.96), and Yellow (0.99). This comparison allows us to understand the proportion of water being stored and released through rivers, given the water surplus derived from the diagnostic net water flux (DNWF > 0).
The median, 25th percentile, 75th percentile, the minimum, and maximum of ratios of P − ET, storage filling based on dS/dt, and river discharge to water surplus derived by the diagnostic net water flux (DNWF > 0) over the TP, Inner basin, and four subbasins.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
Figure 4 also presents ratios of storage filling and river discharge to the diagnostic net water flux (DNWF > 0, water surplus only). As the lengths of data records for dS/dt and Rd are different, independent samples are used in the estimation. The ratio of storage filling (dS/dt > 0) over the TP is about 0.36, while ratios are similar over subbasins of Yangtze (0.42) and Yellow (0.36). With samples of the same time, ratios of river discharge are about 0.40 and 0.36 over Yangtze and Yellow, respectively (figure not shown). These findings suggest that about 20%–30% of water surplus is lost in the form of groundwater discharge and surface flow outside river channels. With all records at Yangtze and Yellow, ratios of river discharge to water surplus are about 0.38 and 0.5. Thus, it is evident that flow outside river channels is an important component in the hydrological cycle.
The ratios of storage filling and river discharge to water surplus presented in Fig. 4 indicate that over half of total samples exceed 0.9 in the subbasin of Brahmaputra. When considering only samples with storage records, the median ratio of river discharge is still high at about 0.64, suggesting that the lateral water inflow plays an essential role in balancing the water budget. In contrast, the median ratio of storage filling is approximately 0.79 in the Inner basin, indicating the existence of lateral water outflow in the endorheic basin. More than 30% of the samples with ratios larger than 1 imply occurrences of lateral water inflow into the Inner basin. In the subbasin of Amu Darya, the water budget is nearly balanced, with the median ratios of the storage filling and river discharge of approximately 0.67 and 0.31, respectively.
The diagnostic net water flux is a simple metric for assessing water surplus at a regional scale, particularly when broadscale observations are lacking over the TP. By combining the net water flux (P − ET referring to DNWF hereafter) with GRACE dS/dt data through a data-driven approach, we can gain valuable insights into how topography and geology impact the water cycle and contribute to changes in hydrological fluxes across the TP and its basins.
b. Seasonality of the recharge–storage–runoff and buffer capacity
The phenomenon of hysteresis is a unique characteristic of natural systems involving recharge, storage and runoff. The periodic behavior observed in Fig. 2 may be influenced by various climatic and physiographic conditions. Figure 5 presents the monthly means of the diagnostic net water flux derived from ERA5, dS/dt from GRACE, runoff estimation based on the water balance between ERA5 and GRACE, and river discharge during 2002–16. The recharge–storage–runoff process is characterized by different timing and amount over the TP, Inner basin, subbasins of Amu Darya, Brahmaputra, Yangtze, and Yellow. In general, the recharge of coupled storages leads to an increase in water mass and runoff, which is referred to as the fast response. In most basins of the TP, the fast runoff response to water surplus (DNWF > 0) is dominated by the boreal summer rainfall.
Monthly mean of terrestrial water balance components (the net water flux DNWF derived from ERA5, dS/dt from GRACE, runoff, and observed river discharge Rd; mm month−1) during 2002–16 over the TP, Inner basin, and four subbasins (Gueymard et al. 2019). Over four subbasins with Rd observations, ratios to the runoff estimation (1.23, 0.86, 0.51, and 0.47) are displayed.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The slow runoff response determined by the storage release is quantified. After the wet season, water mass loss can be supposed to be drained by runoff and evapotranspiration processes (Wu et al. 2021; Kuppel et al. 2017). The runoff process dominates during the dry season over the TP and subbasins of Brahmaputra, Yangtze, and Yellow. Evapotranspiration process also plays a significant role over the Inner basin. The slow response is associated with snowfall in the cold season, as observed over Amu Darya.
The fast response is characterized by the ratio of runoff to net water flux (r1) during the storage filling phase (dS/dt > 0), while the slow response is determined by the ratio of storage release to runoff (r2). Figure 6 presents r1 and r2 during 2002–21 over the TP. The results demonstrate that the fast (slow) response is related to the wet (dry) season, respectively. Storage filling (dS/dt > 0) is predominant from February to August, during which r1 increases from 0.39 to approximately 0.80. The fast runoff response constitutes less than half percent of water surplus associated with snowfall in the cold season and more than 61% during the warm season (June–August) driven by rainfall. In contrast, storage release prevails from September to the following January, consistent with the drying season and the slow runoff response. The ratio of r2 rises rapidly from September, reaching a maximum of 0.87 in November and December before slightly decreasing afterward. r2 occasionally exceeds 1, indicating atmospheric water demand during the dry season.
The dS/dt represents storage filling (cool colors) and release conditions (warm colors) over the TP during 2002–21. The r1 (ratios of runoff to the net water flux with storage filling condition, the positive) and r2 (ratios of storage release to runoff, the negative) correspond to the fast and slow responses, respectively. Multiyear averaged r1 and r2 values during January–December are displayed at the bottom.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The fast and slow runoff responses exhibit distinct characteristics across basins of the TP. In the subbasin of Amu Darya, snowfall dominates from November to the following April with ratios of runoff to net water flux (r1) less than 0.50 (in Fig. 7). Ratios of storage release to runoff (r2) are over 1.50 from June to September, indicating that about one-third to one-half of storage release returned to the atmosphere during the warm season on average, which may impact summer rainfall over the TP. Amu Darya exhibits high runoff-buffering potential, as indicated by the relatively small r1 and large r2, suggesting a small proportion of runoff to the net water flux and a large contribution of storage release to runoff.
As in Fig. 6, but over the subbasin of Amu Darya.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The same analysis is applied to the subbasin of Brahmaputra. However, since r1 and r2 only consider runoff outflow response, the prevailing runoff inflow (R < 0) during April–June resulted in missing values in Fig. 8. The runoff inflow contributes to mass increase (dS/dt > 0) in the first half of the year. The ratios of runoff outflow to the net water flux are predominant only in July and August, with r1 at about 0.50. The ratios of storage release to runoff outflow (r2) exceed 1 on average from October to the following March. During the dry season, no more than one-sixth of storage release is returned to the atmosphere.
As in Fig. 6, but over the subbasin of Brahmaputra.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The characteristics of the Inner basin’s fast and slow runoff responses are shown in Fig. 9. The storage filling period is relatively short, occurring during June–August with r1 at about 0.55. r2 is close to 1 in September, but quickly increases to its maximum (1.76) in October and remains high until the following May. This large r2 indicates that the Inner basin has a high buffering potential to provide runoff and atmospheric water during the dry season. Furthermore, runoff inflow plays a crucial role in terrestrial water increase during the dry season. In particular, storage filling derived by runoff inflow is evident in February. As an endorheic basin, the contribution of runoff through underground passages to the water cycle over the Inner basin cannot be overlooked (Yong et al. 2021).
As in Fig. 6, but over the Inner basin.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
Relatively small r1 and large r2 correspond to high runoff-buffering potential, and vice versa. Figure 10 compares r1 and r2 across the TP, 12 basins and 4 subbasins (shown in Fig. 1). Based on r1 and r2 for the TP (0.61 and 0.5, respectively), the buffer capacity of Amu Darya, Tarim, Indus, and Inner is relatively high due to their smaller r1 and larger r2. However, the slow runoff response is accompanied by atmospheric water demand during the dry season, except for Indus. The buffer capacity is low for Ganges and Salween basins, where about 70% water surplus rapidly drains away during the rainy season.
Averaged r1 (x axis) and r2 (y axis) in the TP, 12 drainage basins, and 4 subbasins (shown in Fig. 1).
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The buffer capacity of a basin is determined by various factors including climatic and terrain drivers. r1 and r2 are indicators derived from reanalysis and satellite datasets. Although they may not be easily quantified as functions of these factors, specific characteristics of buffer capacity can be identified. For example, the subbasin of Brahmaputra is located within the TP and has a higher buffer capacity (r1 of 0.47 and r2 of 0.92) compared to the whole basin (r1 of 0.62). This is due to the deep slopes in the south edge of the TP where rapid runoff is predominant. Similarly, the Inner basin, also located within the TP, has a higher buffer capacity due to its specific climatic and terrain features. In Amu Darya and Indus, relatively small r1 is a result of the prevailing snowfall which is stable enough to store. The presence of glaciers in Amu Darya, Indus, Tarim, and the subbasin of Brahmaputra also contributes to the buffer capacity. Ratios of r2 greater than 1 are associated with dry air conditions. Storage release contributes to evapotranspiration and runoff during the dry season. Therefore, it can be concluded that the buffer capacity is influenced by a complex interplay of various climatic and terrain factors.
c. Long-term trends in river discharge and estimation
On average, the ratio of river discharge to runoff estimation is 1.23 over the subbasin of Amu Darya (shown in Fig. 5c). River discharge is much larger than runoff derived from the water balance during June–September. This indicates an imbalance and water loss due to runoff uncoupled recharge and storage, considering the distribution of snow cover and glaciers over Amu Darya. It may suggest the adverse impact of warming trend over the TP.
Figure 11 compares observed river discharge and runoff estimation, dS/dt, and the net water flux (mm month−1) over the subbasin of Amu Darya during June–September from 2002 to 2021. The river discharge is distinguished from the estimation, suggestive of different mechanisms for their changes. Interannual variation of river discharge is very small, unlike the estimation. The river discharge has a downward trend, while the estimation shows an increasing trend. Observations are roughly twice that of the estimation in the first few years, but since 2012, they have been at a similar level. The high correlation of runoff estimation with dS/dt (−0.75) and the net water flux (−0.63) suggests that the interannual variation of runoff is driven by year-to-year variations of storage release and atmospheric absorption. The increasing trend of runoff estimation is largely attributed to a decreasing tendency of atmospheric water demand. This comparison underscores the importance of accurately estimating runoff to better manage water resources in Amu Darya.
Monthly mean of observed river discharge (Rd, green line) and runoff estimation (R, blue line), water storage release (−dS/dt, red line), and the atmospheric net water flux uptaking (−DNWF, black line) over the subbasin of Amu Darya during June–September from 2002 to 2021. Units are mm month−1.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The decreasing trend in river discharge over Amu Darya (Fig. 11) is likely due to energy constraints rather than water supply limitations. There is almost no correlation (0.03) between river discharge and the net water flux in a longer period of 1979–2017. Figure 12a illustrates the ratio of river discharge to net water flux over the subbasin of Amu Darya, which reveals a significant water loss imbalance particularly in the first decade of the twenty-first century. However, since 2012, the ratio has recovered to levels similar to the 1980s and is close to achieving water balance. Correlations between the ratio and river discharge (0.29) and the net water flux (−0.88) indicate that water imbalance is primarily due to an unusually small net water flux.
Ratios of river discharge to the net water flux over subbasins of Amu Darya, Brahmaputra, Yangtze, and Yellow from 1979.
Citation: Journal of Hydrometeorology 24, 10; 10.1175/JHM-D-23-0045.1
The ratios of river discharge to runoff estimation for the subbasins are 0.86 for Brahmaputra, 0.51 for Yangtze, and 0.47 for Yellow on average (shown in Figs. 5d–f). Since summer rainfall is the primary source of water surplus, river discharge is highly correlated with the net water flux over these subbasins (0.84, 0.67, and 0.60). The ratios of river discharge to the net water flux exhibit distinct features of long-term variability. The downward trend of the ratio from 1979 to 2014 (Fig. 12b) suggests a decrease in the risk of water loss in the subbasin of Brahmaputra. The significant correlation of the ratio with the net water flux (−0.56) indicates the role of increased water supply. Ratios in the subbasins of Yangtze and Yellow have shifted from a downward trend to an upward trend (Figs. 12c,d). Significant correlations of the ratio with river discharge (0.72 and 0.75) confirm decadal variability in river discharge rather than the net water flux. As ratios to water surplus are close to 1 or have increased in the last decade, river discharge in the subbasins of Brahmaputra, Yangtze and Yellow need be carefully monitored. Long-term variability of river discharge over the Yangtze and Yellow may be related to changes in the phase of precipitation, rather than the amount of water supply (Wang et al. 2022), which should be examined more closely in future studies.
4. Discussion
TP is a vital and vulnerable water tower that supports the livelihoods of billions of people (Yao et al. 2022). Given the ongoing climate changes, the hydrological impacts have become a major concern for both scientists and policymakers. Accurate information on water resources at varied temporal and spatial scales is crucial for sustaining ecosystems and ensuring a stable water supply. Unfortunately, obtaining such observations, including precipitation, evapotranspiration, and river discharge, is challenging due to economic and technical constraints, particularly in remote and mountainous regions like the TP (Wang et al. 2021).
The data-driven recharge–storage–runoff perspective provides a useful tool for estimating freshwater flows across watersheds, without addressing the impact of catchment properties on hydrodynamic forces (Sproles et al. 2015; Kuppel et al. 2017; Du et al. 2020). In this study, we use a combination of land–atmosphere water balance, net water flux from ERA5, dS/dt from GRACE and GRACE Follow-On, and runoff estimations to characterize the large-scale water budget at monthly intervals. By comparing runoff estimations to observed river discharge, we aim to gain insight into their potential application for water resources assessment in ungauged areas.
Accurately understanding changes in the recharge–storage–runoff process over the TP depends on reliable and comprehensive data sources. Data uncertainties are considered in the analysis. Although previous studies (Jiang et al. 2021; Yuan et al. 2021) found overestimation of precipitation amount, they highlighted the superior temporal variabilities and spatiotemporal patterns of precipitation in ERA5 compared to the Global Land Data Assimilation Systems (GLDAS) and the Tropical Rainfall Measuring Mission (TRMM) over the TP. To estimate recharge water (P − ET > 0) over the TP, the diagnostic net water flux which incorporates a mass-consistent wind adjustment and a barotropic correction based on ERA5 data (Mayer et al. 2021) is used in this study. Temporal variabilities of the prognostic and diagnostic net water flux are the same, while the diagnostic net water flux plays an important role in mitigating overestimation of the net water flux derived from P and ET over the TP (Figs. 2 and 3). The annual mean of the diagnostic net water flux is 326 mm yr−1, which is a 21% decrease compared to the prognostic net water flux of 412 mm yr−1. Based on the prognostic net water flux accounting for 51.7% of the annual precipitation in ERA5, the DNWF constrains P with 629 mm yr−1.
Precipitation derived from TRMM is 484 mm yr−1 over the TP. ET monitoring by a combined model (ETMonitor driven by multisource satellite observations) is 336 mm yr−1 (Zheng et al. 2022). Compared to ERA5, there are disproportionate reductions in observation-based estimations of P and ET. Since the reduction in observed precipitation is more significant, the net water flux is reduced to only 148 mm yr−1. Analysis of the recharge–storage–runoff process (same as Fig. 5a, but not shown) shows that TRMM precipitation is underestimated and insufficient to meet the increase in terrestrial water storage especially in May and June over TP. Overestimated P in most reanalyses (NCEP1, NCEP2, ERA-Interim, ERA-40, MERRA, and CFSR) had been reported and biases over TP are not equal (You et al. 2015). ERA5 is currently the most reasonable dataset choice.
Daily TWSA data from ITSG-Grace2018 (Eicker et al. 2020) is employed to derive the monthly change in terrestrial water storage (dS/dt) from 2002 to 2021 in this study. The variability in timing and amount of dS/dt confirms the storage dynamic at basin scales in the TP. The annual cycles of dS/dt are approximately in phase with the net water flux (Fig. 5), suggesting a buffer role in holding water and later releasing. TWSA from ITSG-Grace2018 is compared with those from the Center for Space Research at the University of Texas at Austin (CSR; Save et al. 2016) and NASA GSFC GRACE and GRACE-FO MASCON RL06 v1.0 (GSFC; Loomis et al. 2019). Daily ITSG data are averaged on a monthly scale according to the time boundary defined by CSR/GSFC. Correlations over the TP (figure not shown) are 0.77 and 0.78, respectively. Correlations are even higher at basin scales (0.94 and 0.91 for Amu Darya, 0.96 and 0.96 for Brahmaputra, 0.89 and 0.92 for Yangtze, 0.81 and 0.91 for Yellow, 0.86 and 0.81 for Inner). Thus, storage results are not significantly affected by the use of different GRACE products.
Due to the closure achieved with ERA5 net water flux plus GRACE dS/dt, runoff is constrained by land–atmospheric water balance and compared with available river discharge measurements. Recharge (P − ET > 0) of coupled storages leads to increase in water mass and runoff (the fast response). After the wet season, storage release can be drained by runoff (the slow response). To measure the fast and slow runoff responses to net water flux, r1 (the ratio of runoff outflow to the net water flux in storage filling conditions) and r2 (the ratio of storage release to runoff) are constructed. Small proportions of r1 and large contributions of r2 suggest high runoff-buffering potential, and vice versa. Compared with averaged r1 and r2 (0.62, 0.55) at the TP, basins of Amu Darya, Tarim, Inner, and Indus (Ganges and Salween) have higher (lower) buffer capacity (Fig. 10). Long-term changes in buffer capacity have not been found with whole but short period of 2003–15 data (figure not shown).
5. Conclusions
In summary, this study utilizes data-driven recharge–storage–runoff perspective to provide a comprehensive estimation of multiple aspects of the water cycle over the TP and its basins. The key findings are summarized below:
-
The net water flux plays a critical role in determining the recharge over the TP. The annual mean of the diagnostic net water flux (DNWF) is 326 mm yr−1, which is a 21% decrease compared to the prognostic net water flux (P − ET) of 412 mm yr−1. The edge of the TP tends to overestimate P − ET, while the interior underestimates it compared to DNWF. This is attributed to the mass-consistent wind adjustment and a barotropic correction applied to DNWF, which effectively reduces artificial noise over the high topography.
-
Through an analysis of the diagnostic net water flux from ERA5, water storage changes (dS/dt) from GRACE, runoff estimations (R) from the land–atmosphere water balance, and river discharge measurements (Rd), the annual cycle of recharge–storage–runoff are found with distinct features over the TP and basins. Generally, the recharge of coupled storages leads to an increase in water mass and runoff, which is referred to as the fast response and measured by the ratio of runoff to the net water flux (r1). After the wet season, the slow runoff response is determined by the water storage release and measured by the ratio of storage release to runoff (r2). The runoff-buffering potential of a basin is inversely related to r1 and directly related to r2. Over the TP, basins of Amu Darya, Tarim, Inner, and Indus (Ganges and Salween) have higher (lower) buffer capacity.
-
Observed river discharge suggests uncoupled recharge–storage–runoff process at Amu Darya where the decreasing trend in river discharge is likely due to energy constraints rather than water supply limitations.
Acknowledgments.
This paper was jointly funded by the Second Tibetan Plateau Scientific Expedition and Research Program (STEP, Grant 2019QZKK0206) and National Natural Science Foundation of China (Grant 42071408).
Data availability statement.
ERA5 data are developed by ECWMF and supplied by the Climate Data Store (data used here can be found online https://cds.climate.copernicus.eu/). ITSG-Grace2018 data are publicly available from TU Graz (https://www.tugraz.at/institute/ifg/downloads/gravity-field-models/itsg-grace2018).
REFERENCES
Chen, D. L., and Coauthors, 2015: Assessment of past, present and future environmental changes on the Tibetan Plateau (in Chinese). Chin. Sci. Bull., 60, 3025–3035, https://doi.org/10.1360/N972014-01370.
Dai, A., 2016: Historical and future changes in streamflow and continental runoff. Terrestrial Water Cycle and Climate Change: Natural and Human Induced Impacts, Geophys. Monogr., Vol. 221, Amer. Geophys. Union, 17–37, https://doi.org/10.1002/9781118971772.ch2.
Du, H., H. S. Fok, Y. Chen, and Z. Ma, 2020: Characterization of the recharge-storage-runoff process of the Yangtze river source region under climate change. Water, 12, 1940, https://doi.org/10.3390/w12071940.
Duan, A., and Z. Xiao, 2015: Does the climate warming hiatus exist over the Tibetan Plateau? Sci. Rep., 5, 13711, https://doi.org/10.1038/srep13711.
Eicker, A., L. Jensen, V. Wöhnke, H. Dobslaw, A. Kvas, T. Mayer-Gürr, and R. Dill, 2020: Daily GRACE satellite data evaluate short-term hydro-meteorological fluxes from global atmospheric reanalyses. Sci. Rep., 10, 4504, https://doi.org/10.1038/s41598-020-61166-0.
Gueymard, C. A., V. Lara-Fanego, M. Sengupta, and Y. Xie, 2019: Surface albedo and reflectance: Review of definitions, angular and spectral effects, and intercomparison of major data sources in support of advanced solar irradiance modeling over the Americas. Sol. Energy, 182, 194–212, https://doi.org/10.1016/j.solener.2019.02.040.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Immerzeel, W. W., and M. F. P. Bierkens, 2012: Asia’s water balance. Nat. Geosci., 5, 841–842, https://doi.org/10.1038/ngeo1643.
Immerzeel, W. W., L. P. H. van Beek, and M. F. P. Bierkens, 2010: Climate change will affect the Asian water towers. Science, 328, 1382–1385, https://doi.org/10.1126/science.1183188.
Immerzeel, W. W., and Coauthors, 2020: Importance and vulnerability of the world’s water towers. Nature, 577, 364–369, https://doi.org/10.1038/s41586-019-1822-y.
Jiang, Y., K. Yang, C. Shao, X. Zhou, L. Zhao, Y. Chen, and H. Wu, 2021: A downscaling approach for constructing high-resolution precipitation dataset over the Tibetan Plateau from ERA5 reanalysis. Atmos. Res., 256, 105574, https://doi.org/10.1016/j.atmosres.2021.105574.
Kuppel, S., Y. Fan, and E. G. Jobbágy, 2017: Seasonal hydrologic buffer on continents: Patterns, drivers and ecological benefits. Adv. Water Resour., 102, 178–187, https://doi.org/10.1016/j.advwatres.2017.01.004.
Lei, Y., and Coauthors, 2017: Lake seasonality across the Tibetan Plateau and their varying relationship with regional mass changes and local hydrology. Geophys. Res. Lett., 44, 892–900, https://doi.org/10.1002/2016GL072062.
Lei, Y., J. Shi, C. Xiong, and D. Ji, 2021: Tracking the atmospheric-terrestrial water cycle over the Tibetan Plateau based on ERA5 and GRACE. J. Climate, 34, 6459–6471, https://doi.org/10.1175/JCLI-D-20-0692.1.
Li, X., D. Long, B. R. Scanlon, M. E. Mann, X. Li, F. Tian, Z. Sun, and G. Wang, 2022: Climate change threatens terrestrial water storage over the Tibetan Plateau. Nat. Climate Change, 12, 801–807, https://doi.org/10.1038/s41558-022-01443-0.
Loomis, B. D., S. B. Luthcke, and T. J. Sabaka, 2019: Regularization and error characterization of GRACE mascons. J. Geod., 93, 1381–1398, https://doi.org/10.1007/s00190-019-01252-y.
Mayer, J., M. Mayer, and L. Haimberger, 2021: Consistency and homogeneity of atmospheric energy, moisture, and mass budgets in ERA5. J. Climate, 34, 3955–3974, https://doi.org/10.1175/JCLI-D-20-0676.1.
Mayer, M., and L. Haimberger, 2012: Poleward atmospheric energy transports and their variability as evaluated from ECMWF reanalysis data. J. Climate, 25, 734–752, https://doi.org/10.1175/JCLI-D-11-00202.1.
McNamara, J. P., D. Tetzlaff, K. Bishop, C. Soulsby, M. Seyfried, N. E. Peters, B. T. Aulenbach, and R. Hooper, 2011: Storage as a metric of catchment comparison. Hydrol. Processes, 25, 3364–3371, https://doi.org/10.1002/hyp.8113.
Ouyang, X., D. Chen, Y. Feng, and Y. Lei, 2019: Comparison of seasonal surface temperature trend, spatial variability, and elevation dependency from satellite-derived products and numerical simulations over the Tibetan Plateau from 2003 to 2011. Int. J. Remote Sens., 40, 1844–1857, https://doi.org/10.1080/01431161.2018.1482024.
Reeves Eyre, J. E. J., and X. Zeng, 2021: The Amazon water cycle: Perspectives from water budget closure and ocean salinity. J. Climate, 34, 1439–1451, https://doi.org/10.1175/JCLI-D-20-0309.1.
Riegger, J., and M. J. Tourian, 2014: Characterization of runoff-storage relationships by satellite gravimetry and remote sensing. Water Resour. Res., 50, 3444–3466, https://doi.org/10.1002/2013WR013847.
Rodell, M., and J. T. Reager, 2023: Water cycle science enabled by the GRACE and GRACE-FO satellite missions. Nat. Water, 1, 47–59, https://doi.org/10.1038/s44221-022-00005-0.
Save, H., S. Bettadpur, and B. D. Tapley, 2016: High resolution CSR GRACE RL05 mascons. J. Geophys. Res. Solid Earth, 121, 7547–7569, https://doi.org/10.1002/2016JB013007.
Scanlon, B. R., and Coauthors, 2019: Tracking seasonal fluctuations in land water storage using global models and GRACE satellites. Geophys. Res. Lett., 46, 5254–5264, https://doi.org/10.1029/2018GL081836.
Sproles, E. A., S. G. Leibowitz, J. T. Reager, P. J. Wigington Jr., J. S. Famiglietti, and S. D. Patil, 2015: GRACE storage-runoff hystereses reveal the dynamics of regional watersheds. Hydrol. Earth Syst. Sci., 19, 3253–3272, https://doi.org/10.5194/hess-19-3253-2015.
Syed, T. H., J. S. Famiglietti, J. Chen, M. Rodell, S. I. Seneviratne, P. Viterbo, and C. R. Wilson, 2005: Total basin discharge for the Amazon and Mississippi River basins from GRACE and a land-atmosphere water balance. Geophys. Res. Lett., 32, L24404, https://doi.org/10.1029/2005GL024851.
Syed, T. H., J. S. Famiglietti, and D. P. Chambers, 2009: GRACE-based estimates of terrestrial freshwater discharge from basin to continental scales. J. Hydrometeor., 10, 22–40, https://doi.org/10.1175/2008JHM993.1.
Tapley, B. D., and Coauthors, 2019: Contributions of GRACE to understanding climate change. Nat. Climate Change, 5, 358–369, https://doi.org/10.1038/s41558-019-0456-2.
Trenberth, K. E., 1991: Climate diagnostics from global analyses: Conservation of mass in ECMWF analyses. J. Climate, 4, 707–722, https://doi.org/10.1175/1520-0442(1991)004<0707:CDFGAC>2.0.CO;2.
Wang, G., Y. He, J. Huang, X. Guan, X. Wang, H. Hu, S. Wang, and Y. Xie, 2022: The influence of precipitation phase changes on the recharge process of terrestrial water storage in the cold season over the Tibetan Plateau. J. Geophys. Res. Atmos., 127, e2021JD035824, https://doi.org/10.1029/2021JD035824.
Wang, L., and Coauthors, 2021: TP-River: Monitoring and quantifying total river runoff from the third pole. Bull. Amer. Meteor. Soc., 102, E948–E965, https://doi.org/10.1175/BAMS-D-20-0207.1.
Wu, R.-J., M.-H. Lo, and B. R. Scanlon, 2021: The annual cycle of terrestrial water storage anomalies in CMIP6 models evaluated against GRACE data. J. Climate, 34, 8205–8217, https://doi.org/10.1175/JCLI-D-21-0021.1.
Yao, T., and Coauthors, 2012: Different glacier status with atmospheric circulations in Tibetan Plateau and surroundings. Nat. Climate Change, 2, 663–667, https://doi.org/10.1038/nclimate1580.
Yao, T., and Coauthors, 2019: Recent Third Pole’s rapid warming accompanies cryospheric melt and water cycle intensification and interactions between monsoon and environment: Multidisciplinary approach with observations, modeling, and analysis. Bull. Amer. Meteor. Soc., 100, 423–444, https://doi.org/10.1175/BAMS-D-17-0057.1.
Yao, T., and Coauthors, 2022: The imbalance of the Asian water tower. Nat. Rev. Earth Environ., 3, 618–632, https://doi.org/10.1038/s43017-022-00299-4.
Yong, B., C.-Y. Wang, J. Chen, J. Chen, D. A. Barry, T. Wang, and L. Li, 2021: Missing water from the Qiangtang basin on the Tibetan Plateau. Geology, 49, 867–872, https://doi.org/10.1130/G48561.1.
You, Q., J. Min, W. Zhang, N. Pepin, and S. Kang, 2015: Comparison of multiple datasets with gridded precipitation observations over the Tibetan Plateau. Climate Dyn., 45, 791–806, https://doi.org/10.1007/s00382-014-2310-6.
You, Q., and Coauthors, 2020: Elevation dependent warming over the Tibetan Plateau: Patterns, mechanisms and perspectives. Earth-Sci. Rev., 210, 103349, https://doi.org/10.1016/j.earscirev.2020.103349.
Yuan, X., K. Yang, H. Lu, J. He, J. Sun, and Y. Wang, 2021: Characterizing the features of precipitation for the Tibetan Plateau among four gridded datasets: Detection accuracy and spatio-temporal variabilities. Atmos. Res., 264, 105875, https://doi.org/10.1016/j.atmosres.2021.105875.
Zhang, G., T. Yao, H. Xie, S. Kang, and Y. Lei, 2013: Increased mass over the Tibetan Plateau: From lakes or glaciers? Geophys. Res. Lett., 40, 2125–2130, https://doi.org/10.1002/grl.50462.
Zhang, G., and Coauthors, 2017: Lake volume and groundwater storage variations in Tibetan Plateau’s endorheic basin. Geophys. Res. Lett., 44, 5550–5560, https://doi.org/10.1002/2017GL073773.
Zheng, C., L. Jia, and G. Hu, 2022: Global land surface evapotranspiration monitoring by ETMonitor model driven by multi-source satellite earth observations. J. Hydrol., 613, 128444, https://doi.org/10.1016/j.jhydrol.2022.128444.
