Contribution of glacial melt to river runoff as determined by stable isotopes at the source region of the Yangtze River, China

Hydrograph separation is a technique that is used for identifying runoff components, and the derived data are essential for the optimal protection of water resources, for eco-hydrological studies, and for improving flood forecasts (Wels et al. 1991; Wenninger et al. 2004). Both chemical and isotopic tracers have been widely applied in hydrograph separation. Isotopic tracers are assumed to be conservative, while chemical tracers are considered to be quasi-conservative (Richey et al. 1998). Various studies have compared the use of chemical and isotopic tracers and have observed differing results. Some studies found no systematic bias in isotopic or chemical hydrograph separation results (Wels et al. 1991; Leaney et al. 1993; Pionke et al. 1993; Ribolzi et al. 1996), while others observed significant differences between chemical tracer-based and isotope-based hydrograph separation results (Hooper & Shoemaker 1986; Nolan & Hill 1990; Durand et al. 1993; Klaus & McDonnell 2013). These differences are usually attributed to the non-conservative behavior of chemical tracers (Hooper & Shoemaker 1986). In other words, the differences are due to the fact that the reaction rates of chemical tracers can differ in different environments (Wels et al. 1991). Such results show the clear limitations of chemical tracers and have motivated many researchers to use isotopic tracers in hydrograph separation studies (e.g., Klaus & McDonnell 2013).

Different water types in the hydrological cycle have distinct isotopic fingerprints, and therefore, environmental isotopes can be used to identify water sources, study water mass movement, and understand many other hydrological processes (Kendall & Caldwell 1998; Yurtsever 2000). Stable isotopic tracers such as oxygen-18 (¹⁸O) and deuterium (²H) are ideal conservative tracers because they are part of the water molecule and are only subject to changes due to mixing (Klaus & McDonnell 2013). Isotopic tracers have been the most useful tool employed to identify water sources and related information, especially for the identification of runoff components (McGuire & McDonnell 2008; Gimeno et al. 2012).

In general, stable isotope hydrograph separation is a technique based on the assumptions that stream water is a mix of individual water sources and that the specific water sources contributing to river runoff at a particular location are known (Dincer et al. 1970; Maurya et al. 2011). Stable isotopes are generally used for two-component hydrograph separation on an event timescale, while for three-component hydrograph separation, isotopic tracers must be combined with chemical tracers (Ladouche et al. 2001).

Isotopic tracers have been commonly applied to identify event (new) water and pre-event (old) water (e.g., Fritz et al. 1976; Sklash et al. 1976; Sklash & Farvolden 1979; Lakey & Krothe 1996; Ladouche et al. 2001; Huth et al. 2004; Lyon et al. 2009; Zhao et al. 2013; Dahlke et al. 2014) by using the two-component mixing model, which is based on mass balance equations (Pinder & Jones 1969). Some studies have also used this method to separate overland (surface) flow and base (subsurface) flow (Hooper & Shoemaker 1986; Kennedy et al., 1986; Hinton et al., 1994; Soulsby 1995), which are considered similar to event and pre-event water, respectively. For investigating the contribution of snow/glacial melt runoff to the total runoff, the two-component and three-component hydrograph separation methods have both been applied to glacierized regions (Dincer et al. 1970; Shanley et al. 2002; Williams et al. 2006; Maurya et al. 2011; Kong & Pang 2012). Most of these studies were based on individual rainfall or snow/glacial melt events. The isotopic compositions of these individual events exhibit marked temporal variations (Qu et al. 2014). Although there are spatial and temporal variations in the isotopic character of water components including precipitation and streamflow, these variations are reduced when seasonal averages of isotope values over the catchment are used (Maurya et al. 2011). Furthermore, the annual mean contribution of different water sources is more important from the perspective of water resource management and protection than that for an individual event. Therefore, seasonal mean isotopic data is recommended for use in hydrograph separation.

The source region of the Yangtze River (SRYR), which is located in the interior of the Qinghai–Tibet Plateau, is a unique natural environment with abundant natural resources that also serve important ecological functions. The natural surrounding environment and the aquatic ecosystem are both very fragile and linked to water availability. Glacial retreat, albeit with strong regional variability, has been observed on the Qinghai–Tibet Plateau (Yao et al. 2004). While magnitudes have varied, glaciated areas at the SRYR have shrunk during the past four decades (Ye et al. 2006; Zhang et al. 2008; Wu et al. 2013; Yao et al. 2014). Changes to the region's glaciers can ultimately affect the ecosystem as well as have an impact on regional water resources at the SRYR. Therefore, it is essential to investigate the contribution of glaciers to total runoff.

In this study, a seasonal stable isotope dataset (from the observational period 1 June 2013 to 10 June 2014), for precipitation, glacial meltwater, and streamflow was applied to the isotope hydrograph separation method to quantify the contribution of glacial meltwater to total runoff in a high altitude catchment of the Gaerqu River, which is located in the SRYR, China. This study focused on quantifying the contributions of the two different water sources (precipitation and glacial meltwater) identified in the streamflow at the catchment outlet. Soil water, also known as base flow or groundwater, is also a product of precipitation and glacial meltwater. Therefore, soil water movement via hydrological processes was not reflected in the results.

Because of the harsh natural environment, the study area has not been affected much by human activities such as land use change. This was beneficial for our scientific research. In this study, the hydro-meteorological variables observed included precipitation, air temperature, and streamflow. Water samples of precipitation, glacial melt, and river water were collected and analyzed for stable isotopes. With the results of these analyses, the isotope hydrograph separation method was used to quantify the contribution of glacial melt to total runoff during the entire observational period. Finally, the degree-day method was applied to investigate temporal variations in glacial melt runoff. It is hoped that the results of this study will provide a useful resource to support water resource management and protection efforts, especially as the SRYR undergoes glacial changes.

The SRYR is located in the central eastern part of the Qinghai–Tibet Plateau. With an average altitude of 4,754.5 m above sea level (a.s.l.), this region includes one of the highest major river basins in the world. Glaciers of the SRYR are primarily distributed on the northern slope of the Tanggula Mountains, the southern slope of the Kunlun Mountains, and on Sedir Mountain. According to the Chinese Glacier Inventory (Shi et al. 2005), 753 glaciers are located in the SRYR, and these have a combined total volume in water equivalents of approximately 887.5 × 10⁸ m³, which is five times the annual discharge of the SRYR.

The Gaerqu River is one of main headwater rivers of the SRYR. Its catchment area is 4,482.0 km² and has a mean altitude of ∼5,015.4 m a.s.l. (with a total altitude range of 4,491 to 6,621 m a.s.l). The arid to semi-arid climate of the catchment results from its location and from the topographical features of the Qinghai–Tibet Plateau. Mean annual precipitation is <300 mm, and precipitation falls mainly between June and September coincident with the annual peak of glacial melt for the region. Because of the high altitude, the environment of the catchment is very cold. The mean air temperature between June 2013 and June 2014 was −3.7 °C. The mean values of the daily maximum and minimum air temperatures during the same period were 5.1 °C and −10.0 °C, respectively. Rivers are frozen during the winter season (DJF).

Water samples, including those for precipitation, glacial melt, and river water, were collected during the observational period. Precipitation samples were collected for each precipitation event (including rainfall and snowfall) at the observation sites (Figure 1). Samples of glacial meltwater were collected at the Geladandong Glacier, which supplies runoff to the catchment. The numbers of water samples for precipitation, glacial melt, and river water were 12, 42, and 40, respectively.

Landsat 8 Operational Land Imager (OLI) and Landsat Enhanced Thematic Mapper Plus (ETM+) imagery data were analyzed to detect glacier margins in the catchment. The Landsat imagery data were obtained from http://glovis.usgs.gov/. The images (Path 138 and Row 037) covered the entire catchment. In total, four images were obtained, including two OLI images (May 21, 2013 and September 10, 2013) and two ETM+ images (August 1, 2013 and July 3, 2014).

The stable isotopes of precipitation and river water samples were analyzed using a liquid-water isotope analyzer (Picarro L2120-i) at the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (CAS).

The Picarro L2120-i provides measurements of δ¹⁸O and δ²H in liquid-water samples with an accuracy better than 0.2‰ for ¹⁸O/¹⁶O and 0.6‰ for ²H/¹H. Both δ¹⁸O and δ²H were measured for all water samples with the exception of glacial melt. Glacial meltwater samples were analyzed for δ¹⁸O alone using the Picarro L2120-i at the Institute of Tibetan Plateau Research, CAS.

Glacier margins were identified using remote sensing images. Manual interpretation, which has proved to be the best method for delineating glacier areas from satellite imagery (Bishop et al. 2004; Raup et al. 2007), was selected to delineate glacier boundaries in the catchment. This method is particularly effective when delineation is conducted by the same person using a combination of different types of imagery (Paul et al. 2002). Glacial areas were manually digitized from satellite images using the software ArcGIS 10.0. Glacier outlines were mapped manually by using false color composite bands.

In this study, the unit used for the DDF was m³ °C⁻¹d⁻¹ and not mm°C⁻¹d⁻¹, which has been used often in other studies. This choice was made because the chosen unit is more convenient for calculating glacial melt runoff. Due to data limitations, the revised degree-day method, which requires more datasets (including radiation data), was not used in this study.

Based on the observed air temperature data, the air temperature lapse rate at the Dongkemadi Glacier was identified as 0.63 °C/100 m and 0.67 °C/100 m by Liu et al. (2006) and Qiao et al. (2010), respectively. The Dongkemadi Glacier is located in the Tanggula Mountains of the SRYR, and it is near to the Geladandong Glacier (the glacier used in this study). A mean air temperature lapse rate of 0.65 °C/100 m was used as the air temperature lapse rate for the study area. This value was subsequently used to calculate the air temperature at the glacier margin, as identified from remote sensing imagery.

The observed streamflow was greatest from July to early September coincident with the highest precipitation and highest air temperatures recorded during the hydrological year. The maximum flow during the observation period was 22.9 m³/s, and this occurred on September 8, 2013. Streamflow dropped abruptly in mid-September from a mean of 19.7 m³/s (plentiful flow period: July–early September) to a mean of 12.9 m³/s (normal flow: mid September–November). The river froze in November. This abrupt change was facilitated by the rapid decrease in precipitation levels in the catchment. Despite little difference in precipitation between this normal water flow period and June 2013, the streamflow in June 2013 was clearly larger. This increase was likely driven by the seasonal snowmelt that occurs in May–June, when the mean air temperatures are above 0 °C. This would also explain the peak river flow observed in May 2014, despite little precipitation.

In summary, the hydrograph showed a strong correlation between streamflow and precipitation, with the correlation coefficient (R) being 0.86. However, glacial melt runoff could also explain some of the observed streamflow variations, since there was also a positive relationship between streamflow and the mean air temperature (R = 0.73), and air temperature was positively correlated with glacial melt runoff (R = 0.998).

During the hydrological year, total runoff was ∼2.6 × 10⁸ m³, which is equivalent to 57.0 mm of precipitation. The total precipitation observed during the hydrological year was 243.9 mm, which is equivalent to ∼10.9 × 10⁸ m³. Based on precipitation alone, the runoff coefficient, as defined by the ratio of annual runoff to precipitation, was 0.23 for the catchment. If glacial melt runoff was not included, the hydrological budget equation could be described as that precipitation subtracted by runoff was equal to the sum of evapotranspiration and changes in soil water storage at a catchment. This value suggests that precipitation accounted for <25% of runoff throughout the hydrological year, with the remainder lost to evapotranspiration or to changes in soil water storage in the catchment. However, in reality, the precipitation loss must be >76.6% because the true runoff coefficient reflects the combination of both precipitation and glacier melt.

The line was fitted as δ²H = 7.75 δ¹⁸O + 5.93, with a regression line R² value of 0.98. The slope was close to that of the Tingri station (Kang et al. 2002), which was in the northern slope of the Himalayas. The LMWL of the study area had a smaller slope and intercept as compared to the Global Meteoric Water Line (GMWL, δ²H = 8 δ¹⁸O + 10), as defined by Craig (1961). The catchment is located on the north slope of the Tanggula Mountains, which is a barrier for Indian Ocean moisture transport. Before water vapor transport to the catchment, isotopic fractionation during processes of precipitation and evaporation made losses of heavy isotope (¹⁸O and δ²H). This included the continental effect, the latitude effect, and the altitude effect. These processes likely led to the smaller intercept of the LMWL as compared with the GMWL. The depletion rate of δ²H during isotopic fractionation is larger than that of δ¹⁸O. This resulted in the smaller slope of the LMWL as compared with the GMWL.

During the observation period, the volume-weighted mean values of δ¹⁸O for streamflow at the catchment outlet, for the precipitation at the two sample sites, and for glacial meltwater were −11.9‰, −11.3‰, and −15.6‰, respectively. Calculations using the isotope hydrograph separation method, Equation (4), show that the contribution of glacial melt runoff to total runoff was 15.0%. Uncertainty results for hydrograph separation were shown in Table 1. Calculated by Equation (7), the uncertainty of the separation was ± 3.7% at the confidence level of 95%. Therefore, the contribution of glacial melt runoff was 15.0 ± 3.7% (the confidence level of 95%) at the Gaerqu River catchment. As the total runoff during the observation period was 2.55 × 10⁸ m³, the glacial melt runoff can be calculated as (0.38 ± 0.09) × 10⁸ m³.

The runoff coefficient for the catchment was calculated as 0.23, which indicates that runoff constituted 23% of the total precipitation throughout the study period (a single hydrological year). Since the contribution of glacial melt runoff to total runoff was calculated as 15.0%, the contribution of precipitation to total runoff was 85.0%. Then, runoff coming from precipitation was calculated as the total runoff multiplied by 85.0%. Therefore, precipitation-induced runoff constituted 19.9% of the total annual precipitation, meaning that precipitation loss was <80% across the hydrological year. Precipitation was mainly lost due to evapotranspiration or changes in soil water storage in the catchment.

The LMWL of the Gaerqu River catchment was fitted as δ²H = 7.75 δ¹⁸O + 5.93. This line had a smaller slope and intercept as compared to the GMWL. During the transportation of marine moisture to the catchment, stable isotopes (δ¹⁸O and δ²H) were depleted due to the continental effect, the latitude effect, and the altitude effect. These processes led to the smaller intercept of the LMWL as compared with the GMWL. The depletion rate of δ²H during isotopic fractionation was larger than that of δ¹⁸O, which resulted in the smaller slope of the LMWL as compared with the GMWL. The isotopic composition of stream water is located across the Evaporation Line. The intersection point between the Evaporation Line and the LMWL represented the remaining water on the ground after evaporation. Because of isotopic fractionation in evaporation, the remaining water was enriched with heavy isotopes (¹⁸O and δ²H). In other words, the enrichment of ¹⁸O and δ²H had a positive correlation with evaporation. For the intersection point, the δ¹⁸O and δ²H values were more enriched at downstream water than that at upstream water. This indicated evaporation was greater over the entire catchment than that for the upstream region alone. It is possible that this reflects the lower mean air temperatures of the upstream region.

When applying the isotope hydrograph separation method, the use of mean isotopic values calculated over a long period can help to reduce the uncertainties caused by the lag time of flow concentrations. Applying the isotope hydrograph separation method to precipitation-runoff events is difficult because they are affected by a lag time of flow concentration. However, in this study, river water samples were not collected immediately after every precipitation event, and as a result, isotope hydrograph separations for individual precipitation-runoff events were not included in this study.

Hydrograph separation was focused on the contribution of the initial sources of runoff in the catchment (precipitation and glacial melt) to the total runoff. The isotopes in river water reflect the initial water sources. In this study, isotopic movement through other hydrological processes was not considered, for example, isotope fractionation during the evaporation process.

In this study, it was assumed that glacial melt originates at glacier margins and that the degree-day factor was constant at the glacier margin zone. Although glacier margins were necessarily changing as a result of glacial melt, at the altitude of this study these changes were <15 m during the observation period. Therefore, the degree-day factor at the glacier margin zone, which was identified midway through the observation period, was representative of the glacial melt characteristics across the whole period.

Observed hydro-meteorological data and isotope data of 94 water samples were applied using the isotope hydrograph separation method to quantify the contribution of glacial meltwater to total runoff in the Gaerqu River catchment. The degree-day method was used to detect temporal variations in glacial melt runoff. The contribution of glacial melt runoff to total runoff, as identified by the isotope hydrograph separation method, was 15.0%. Glacial melt runoff was mainly generated in June, July, and August, with the largest monthly value in August 2013 (15.4 × 10⁶ m³). The greatest daily glacial melt runoff recorded was 1.0 × 10⁶ m³, which occurred on August 18, 2013. It is hoped that the results of this study will provide a useful resource to support water resource management and protection efforts, especially as the SRYR undergoes glacial changes.

This study was supported and funded by the National Natural Science Foundation of China (Grant No. 41371058, No. 41201035, and No. 41190080). Isotope data for glacial meltwaters were provided by Prof. Shichang Kang. Thanks are also extended to Lei Wang and Liguang Jiang for assistance in the field.