The dynamics of basin-scale water budgets
over the Tibetan Plateau (TP) are not well understood nowadays due to the lack of in situ hydro-climatic
observations. In this study, we investigate the seasonal cycles and trends of
water budget components (e.g. precipitation
As the highest plateau in the globe (the average elevation is higher than 4000 m above the sea level), the Tibetan Plateau (TP, also called “the roof of the world” or “the third Pole”) is regarded as one of the most vulnerable regions under a warming climate and is exposed to strong interactions among the atmosphere, hydrosphere, biosphere and cryosphere in the earth system (Duan and Wu, 2006; Yao et al., 2012; Liu et al., 2016b). It also serves as the “Asian water tower” from which some major Asian rivers such as the Yellow River, Yangtze River, Brahmaputra River, Mekong River, Indus River, etc., originate, which is a vital water resource to support the livelihoods of hundreds of millions of people in China and the neighbouring Asian countries (Immerzeel et al., 2010; Zhang et al., 2013). Hence, sound knowledge of water budget and hydrological regimes in TP river basins and their responses to the changing environment would have practical relevance for achieving sustainable water resource management and environmental protection in this part of the world (Yang et al., 2014; Chen et al., 2015).
Despite the importance of the TP in this geographic region, advances in
hydrological and land surface studies in this region have been limited by
data scarcity (Zhang et al., 2007; Li et al., 2013; Liu et al., 2017).
For instance, less than 80 observation stations (
Most recently, several global (or regional) datasets relevant to the calculation of water budget have been released. They include remote-sensing-based retrievals (Tapley et al., 2004; Zhang et al., 2010; Long et al., 2014; Zhang et al., 2016), LSM simulations (Rui, 2011), reanalysis outputs (Berrisford et al., 2011; Kobayashi et al., 2015) and gridded forcing data interpolated from the in situ observations (Harris et al., 2014). For example, there are many products related to terrestrial evapotranspiration (ET), such as GLEAM_E (Global Land surface Evaporation: the Amsterdam Methodology, Miralles et al., 2011), MTE_E (a product integrated the point-wise ET observation at FLUXNET sites with geospatial information extracted from surface meteorological observations and remote sensing in a machine-leaning algorithm, Jung et al., 2010), LSM-simulated ETs from Global Land Data Assimilation System version 2 (GLDAS-2) with different land surface schemes (Rodell et al., 2004), ETs from Japanese 55-year reanalysis (JRA55_E), the ERA-Interim global atmospheric reanalysis dataset (ERAI_E), and the National Aeronautic and Space Administration (NASA) Modern Era Retrosphective-analysis for Research and Application (MERRA) reanalysis data (Lucchesi, 2012). Moreover, there are also several global or regional LSM-based runoff simulations from GLDAS and the variable infiltration capacity (VIC) model (Zhang et al., 2014). A few attempts have been made to validate multiple datasets for certain water budget components and to explore their possible hydrological implications. For example, Li et al. (2014) and Liu et al. (2016a) evaluated multiple ET estimates against the water balance method on annual and monthly timescales. Bai et al. (2016) assessed streamflow simulations of GLDAS LSMs in five major rivers over the TP based on the discharge observations. Although uncertainties might exist among different datasets with various spatial and temporal resolutions which are calculated using different algorithms (Xia et al., 2012), they offer an opportunity to examine the general basin-wide water budgets and their uncertainties in gauge-sparse regions such as the TP considered in this study.
From the multiple-dataset perspective, this study aims to investigate the water budget in 18 TP river basins distributed across the Tibetan Plateau and evaluate seasonal cycles and annual trends of these water budget components. This paper is organized as follows: the datasets and methods applied in this study are described in Sect. 2. The results of seasonal cycles and annual trends of water budget components for the river basins are presented and discussed in Sect. 3. The uncertainties arising from employing multiple datasets are also discussed in the same section. In Sect. 4, we generalize our findings, which could be helpful for understanding the water balances of the river basins under constant influence of interplay between westerlies and monsoons (e.g. Indian monsoon, east Asian monsoon) in the Tibetan Plateau.
Overview of multi-source datasets applied in this study.
We obtained the observed daily runoff (
In terms of precipitation (
To get the change in terrestrial storage (
We obtained the monthly gridded temperature dataset (0.5
To quantify the dynamics of vegetation of each river basin, we applied the
normalized difference vegetation index (NDVI) and the leaf area index (LAI; Table 1). Briefly, the NDVI data were obtained from the Global Inventory
Modeling and Mapping Studies (GIMMS; Turker et al., 2005;
In general, the TP climate is under the influences of the westerlies, Indian
summer monsoon and east Asian summer monsoon (Yao et al., 2012). To
investigate the changes in monsoon systems and their potential impacts on
water budgets in the TP basins, we used three monsoon indices, namely Asian
zonal circulation index (AZCI), Indian Ocean dipole mode index (IODMI) and
east Asian summer monsoon index (EASMI). Briefly, the IODMI (reflects the
dynamics of Indian summer monsoon) is an indicator of the east–west
temperature gradient across the tropical Indian Ocean (Saji et al., 1999),
which can be downloaded from the following website:
Main features of the 18 TP river basins used in this study. The precipitation and temperature statistics for each basin were calculated from the observed CMA datasets while the NDVI and LAI statistics were extracted from the GIMMS NDVI dataset and GLASS LAI product. The GA and SC percentages represent the percentages of multiyear-mean glacier cover and snow cover in each basin, which were calculated from the second glacier inventory dataset of China and the daily TP snow composite products (2005–2013).
Map of river basins and hydrological gauging stations (green dots) over the Tibetan Plateau (TP) used in this study. The grey shading shows the topography of the TP in metres above sea level, and the blue shading exhibits the glaciers distribution in the TP extracted from the second glacier inventory dataset of China.
In this study, we selected 18 river basins of varied sizes (range:
2832–191 235 km
The basin-wide water balance on the monthly and annual timescales could be
written as the principle of mass conservation (also known as the continuity
equation, Oliveira et al., 2014) of basin-wide precipitation (
The Mann–Kendall (MK) test is a rank-based non-parametric approach which is
less sensitive to outliers relative to other parametric statistics, but it is
sometimes influenced by the serial correlation of time series. Pre-whitening
is often used to eliminate the influence of lag-1 autocorrelation before the
use of MK test. For example,
Annual-averaged water storage changes (
We first assessed the VIC_IGSNRR-simulated runoff against the
observations for each basin (for example, at Tangnaihai and Pangduo stations
in Fig. 2). If the Nash–Sutcliffe efficiency (NSE) coefficient between the observation
and simulation is above 0.65, the VIC_IGSNRR-simulated runoff
is acceptable and could be used to replace the missing runoff values for a
given basin. Moreover, the CMA precipitation is consistent with TRMM (correlation coefficient (corr.)
Comparison of VIC_IGSNRR-simulated and observed
monthly runoff for Tangnaihai and Panduo stations
We then evaluated 6 ET products in 18 TP basins against our
calculated
Comparison of different ET products against the calculated ET
through the water balance method (ET
To investigate the general hydro-climatic characteristics of river basins
over the TP, we classify 18 basins into 3 categories, namely
westerlies-dominated basins (Yerqiang, Yulongkashi and Kelia), Indian-monsoon-dominated basins (Brahmaputra and Salween), and east-Asian-monsoon-dominated basins (Yellow, Yalong and Yangtze) according to Tian et al. (2007), Yao et al. (2012) and Dong et al. (2016). Interestingly, they
are clustered into three groups under Budyko framework (Budyko, 1974; Zhang
et al., 2016) with a relatively lower evaporative index in Indian-monsoon-dominant basins and a higher aridity index in westerlies-dominant
basins, which reveal various long-term hydro-climatologic conditions (Fig. 4). Overall, in the westerlies-dominant, Indian-monsoon-dominant and east Asian monsoon-dominant basins, the annual mean air temperature (
General water and energy status (
The multi-year means of water budget components (i.e.
Seasonal cycles (1982–2011) of water budget components in westerlies-dominated (row 1), east-Asian-monsoon-dominated (rows 2–4) and Indian-monsoon-dominated (rows 5–6) TP basins.
Although the temporal patterns of hydrological components are generally analogous, they vary among the parameters, climate zones and even basins (Zhou et al., 2005). For example, relative to air temperature, the seasonal pattern of runoff is similar to precipitation, which reveals that runoff is mainly controlled by precipitation in most TP basins. It is in agreement with that summarized by Cuo et al. (2014). In the westerlies-dominated basins, the peak values of precipitation and runoff are mainly concentrated in June–August, which contribute approximately 68–82 and 67–78 % of annual totals, respectively. During this period, the runoff always exceeds precipitation, which indicates large contributions of glacier and snow meltwater to streamflow. It is consistent with the existing findings in the Tarim River (Yerqiang, Yulongkashi and Keliya rivers are the major tributaries of Tarim River), which indicated that the meltwater accounted for about half of the annual total streamflow (Fu et al., 2008). The ET (vegetation cover) values in three westerlies-dominated basins are relatively smaller (scarcer) than that in other TP basins while the percentages of glacier and seasonal snow cover are higher in these basins which contribute more meltwater to river discharge (Figs. 6 and 7). Overall, the SWE in Yerqiang, Yulongkashi and Keliya rivers are higher in winter than other seasons, but they vary with basins and products which reflect considerable uncertainties in SWE estimations.
Seasonal cycles (1982–2011) of air temperature and vegetation parameters in westerlies-dominated (row 1), east-Asian-monsoon-dominated (rows 2–4) and Indian-monsoon-dominated (rows 5–6) TP basins.
In the Indian-monsoon- and east-Asian-monsoon-dominated basins, the runoff is concentrated during June–September (or June–October), with precipitation being the dominant contributor of annual total runoff. For example, the peak values of precipitation and runoff occur during June-September at Zhimenda station (contributing about 80 and 74 % of the annual totals) while those occur during June–October at Tangnaihai station (contributing about 78 and 71 % of the annual totals, respectively). The results are quite similar to the related studies in the eastern and southern TP such as Liu (1999), Dong et al. (2007), Zhu et al. (2011), Zhang et al. (2013) and Cuo et al. (2014). The vegetation cover (ET) in most basins is denser (higher) than that in the westerlies-dominant basins. Moreover, the seasonal snow mainly covers from mid-autumn to spring and correspondingly the SWE is relatively higher in these months in all basins except for the Yellow River above Xining station, Salwee River above Jiayuqiao station and Brahmaputra River above Nuxia and Yangcun stations.
Seasonal cycles (1982–2011) of snow cover and snow water equivalent (SWE) in westerlies-dominated (row 1), east-Asian-monsoon-dominated (rows 2–4) and Indian-monsoon-dominated (rows 5–6) TP basins. The snow cover was extracted from a cloud-free snow composite product during the period 2005–2013. It should also be noted that the GlobSnow data are not available for some basins.
The
Sen's slopes of water budget components and vegetation parameters
in westerlies-dominated TP basins during the period of 1982–2011. To clearly
exhibit the non-parametric trends of all variables in one panel, the Sen's
Slopes of
Linear and non-parametric trends of westerlies, the Indian monsoon and the east Asian summer monsoon during the period 1982–2011, revealed prospectively by the Asian zonal circulation index, Indian Ocean dipole mode index and east Asian summer monsoon index.
In the east-Asian-monsoon-dominated basins, there are two types of change
for basin-wide water budget components. For example,
Similar to Fig. 8 but for east-Asian-monsoon-dominated TP basins. It should be noted that the GlobSnow data are not available for some basins. The double red stars showed that the trend was statistically significant at the 0.05 level.
The
Similar to Fig. 8 but for Indian-monsoon-dominated TP basins. It should be noted that the GlobSnow data are not available for some basins. The double red stars showed that the trend was statistically significant at the 0.05 level.
Non-parametric trends for different ET estimates during the period 1982–2006, detected by modified Mann–Kendall test. The bold numbers show the detected trend is statistically significant at the 0.05 level.
The results may be unavoidably associated with some uncertainties inherited from
the multi-source datasets used. The primary sources of uncertainty may arise
from the precipitation inputs. We compared the seasonal cycles and annual
trends in different precipitation products, i.e. CMA_P,
IGSNRR_P and TRMM_P (and their
calculated
Uncertainties in seasonal cycles of ETwb calculated from three precipitation products (CMA gridded, IGSNRR_Forcing and TRMM precipitation) in 18 TP basins. The comparisons were conducted during the period 2000–2011 when TRMM data were available.
Although the seasonal cycles of
Uncertainties in annual trends of ET
The interpolation of missing values of runoff with VIC_IGSNRR-simulated runoff, and the gridded precipitation data (which interpolated from
limited gauged precipitation over the plateau) also introduced
uncertainties. There are also considerable uncertainties arising from the process of extending the ET series back empirically prior to the GRACE era. However, the
trends in
In this study, we investigated the seasonal cycles and trends of water
budget components in 18 TP basins during the period 1982–2011, which is not
well understood so far due to the lack of adequate observations in the harsh
environment, through integrating the multi-source global and regional datasets
such as gauge data, satellite remote sensing and land surface model
simulations. By using a two-step bias-correction procedure, we calculated
the annual basin-wide
From the Budyko framework perspective, the general water and energy budgets are different in the westerlies-dominated (with higher aridity index, runoff coefficient and glacier cover), the Indian-monsoon-dominated and the east-Asian-monsoon-dominated (with higher air temperature, vegetation cover and evapotranspiration) basins. In the 18 TP basins, precipitation is the major contributor to the river runoff, which is concentrated mainly during June–October (June–August for the westerlies-dominated basins, June–September or June–October for the Indian-monsoon-dominated and the east-Asian-monsoon-dominated basins). The basin-wide SWE is relatively high from mid-autumn to spring for all 18 TP basins except for Keliya River and Brahmaputra River above the Nuxia and Yangcun stations. There is relatively less vegetation cover, whereas there is more snow–glacier cover in the westerlies-dominant basins compared to other basins.
During the period 1982–2011,
The basin-wide water budget series in the TP rivers used in this study are available from the authors upon request (liuwb@igsnrr.ac.cn).
WL and FS developed the idea to see the general water budgets in the TP basins from the perspective of multi-source datasets. WL collected and processed the multiple datasets with the help of YL, GZ, WHL, HW and PB and prepared the paper. The results were extensively commented on and discussed by FS, JL and YFS.
The authors declare that they have no conflict of interest.
This study was supported by the National Key Research and Development Program of China (2016YFC0401401 and 2016YFA0602402), National Natural Science Foundation of China (41401037, 41601035, 91647110, 41701019 and 41330529), the Open Research Fund of State Key Laboratory of Desert and Oasis Ecology in Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences (CAS), the Key Research Program of the CAS (ZDRW-ZS-2017-3-1), the CAS Pioneer Hundred Talents Program (Fubao Sun), the CAS President's International Fellowship Initiative (2017PC0068), and the program for the “Bingwei” Excellent Talents from the Institute of Geographic Sciences and Natural Resources Research, CAS. We are grateful to the NASA MEaSUREs Program (Sean Swenson) for providing the GRACE land data processing algorithm. We thank Axel Kleidon, the editors and the reviewers for their invaluable comments and constructive suggestions. Edited by: Ross Woods Reviewed by: three anonymous referees